FORUM DISRUPTION: Server Upgrade

[EEVblog]

The server will get more RAM tomorrow and will go down for up to a few hours:
https://www.timeanddate.com/worldclock/fixedtime.html?msg=EEVblog+Server+down+for+upgrade&iso=20170328T02&p1=70&ah=2

[gnif]

Please also note that after the server comes back online there will be a period of tuning which will cause various errors and timeouts, I will update this thread when this is completed, there is no need to report these and I apologize in advance for the frustration it will cause.

[gnif]

The server has been configured to take advantage of the additional RAM, it will be monitored closely over the next few days to see if any additional improvements can be made. There should be no more outages unless noted here prior. Performance feedback is very welcome here.

[mrpackethead]

Am getting the 502 bad gateway errors again just a few minutes ago, though the forum is going just fine. 

[EEVblog]

Was supposed to get 16GB RAM, but got 32GB for some reason 

The MySQL queries has jumped fairly dramatically, seems it now able to process many more requests per second 

[The Soulman]

Thanks dudes! 

[FrankBuss]

It feels a lot faster now when opening topics. Before the upgrade sometimes I had to wait a few seconds, couldn't see this now while reading some topics.

[EEVblog]

Seems faster for me too.
Opening topics, opening a reply, and posting a reply.

[nugglix]

Thread refresh (F5 in open thread page) is now in the sub-second range.
Clearly some sort of success, I'd say. 

Thanks for the time/money spent!

Hmm... question: do I need to read faster now?   

PS: Posting preview also sub-second response.

[MrBungle]

When you said tomorrow, I figured tomorrow.... also known as Wednesday.
All good, minimal disruption 
Great work guys.

[boffin]

After you've run for a couple of days, I highly recommend mysqltuner.pl ; it'll help show bottlenecks (and over-allocations) in your existing my.cnf

[gnif]

After you've run for a couple of days, I highly recommend mysqltuner.pl ; it'll help show bottlenecks (and over-allocations) in your existing my.cnf

It is the tool of the trade

[mikeselectricstuff]

The slowness I've been seing over the last week or so seems to have gone away.

[JPortici]

Browsing o  tapatalk... super fast

[Muttley Snickers]

In gnif we trust, well done and many thanks.   

[gnif]

The number of update and insert requests seems quite stable over the past month. But the number of select requests has increased markedly. Why is that? It predates todays RAM upgrade. IIRC you re-jigged the MySQL tables. Is the increase simply because MySQL does more selects as a result of the DB work? Does it reflect new selects being generated as a result of a similar load?
Were you RAM constrained before and did that contribute to the record locking for long periods during maintenance? Which lead to the 502's. Just reading what has been reported I don't quite see how more RAM has addressed the problems reported. Were there performance issue at times outside the DB maintenance intervals?

Sorry gnif I don't mean to bombard you with questions it is just that I am not sure of the right question in MySQL speak.

No worries mate. The pickup in qps is coincides with the MyIASM to InnoDB conversion on those final few tables, we are seeing far fewer lock contentions and as such the total server throughput went up. It went up because prior when a table was locked, it would put the next query into a queue, this is fine provided the queue length doesn't blow out, like it started to. In short, we were limited by locking, which throttled throughput.

Once we switched to row level locking, there is no longer (much) lock contention, and as such all these requests now can run concurrently, so this increases our throughput, people get faster response, etc... This explains the jump in queries per second. I also went through and added some indexes to fix some slow queries, SMF developers don't seem to understand the importance of them, this is not the first time I have had to fix queries.

But then we hit I/O issues as we had also increased our IOPS, so instead of seeing as many 502's due to PHP being too busy, we were seeing more slower responses due to the now much increased IO load, which in turn affects the entire server, not just DB requests. Most people throw a SSD in to resolve this, but it is unwarranted in most cases and at best a SSD is a temporary fix. The usually better solution is to throw a bit more ram at the server and cache the indexes and entire working data set, since RAM is orders of a magnitude faster then a SSD, and once the server has warmed up (data is in RAM), it barely touches the disk except for updates/inserts.

Finally since we have more RAM then we expected, I thew some extra to Memcached.

[rrinker]

 I hadn't really noticed it being slower than normal, and just had a bunch of 502's but that was during the time gnif said it was rebuilding the database after the previous changes - that went away.

 But wow - compared to yesterday, it is flying now. Incredibly smooth and responsive clicking on messages and going through pages.    More RAM, more gooder.

[timgiles]

Yes - buttery smooth today 

[grumpydoc]

Very noticeable improvement!

 

[2N3055]

Page loading is different: slight delay and than faster loading.. All in all a bit better.. But biggest difference is when replying.. Posting goes much faster....

[dr.diesel]

Serious improvement from the sticks of Indiana as well, big thanks to gnif/Dave.

[tautech]

Big  here too. 

[mrpackethead]

Are you serving pages on ipv6 yet?

[gnif]

Are you serving pages on ipv6 yet?

Unfortunately not, sadly IPv6 has not been widely deployed by data centers.

[EEVblog]

The forum is seriously slick from my lab too. So both using the new fixed IP trial setting, and without.
That RAM really did the trick.
BTW, I still don't know what it's going to cost me.
The head of marketing at HostGator (who is a fan BTW) said they'll sort it out after I complained about the ludicrous price that would have cost an extra $240/month to go from the 8GB we had to 32GB we have now. PER MONTH!

[FrankBuss]

My hosting provider, 1and1, has full dedicated servers with 32 GB RAM, 12 cores x 2.3 GHz and 2 TB RAID 1 harddisk for $200 per month. Just in case you need some arguments for HostGator

[Rerouter]

Isn't that the point where you send off ram sticks in the post ?

[gnif]

My hosting provider, 1and1, has full dedicated servers with 32 GB RAM, 12 cores x 2.3 GHz and 2 TB RAID 1 harddisk for $200 per month. Just in case you need some arguments for HostGator

I am paying $130 for E3-1245, 32GB RAM and 2x 480GB SSD... and that is in Australia! Why are HG so expensive?

[EEVblog]

My hosting provider, 1and1, has full dedicated servers with 32 GB RAM, 12 cores x 2.3 GHz and 2 TB RAID 1 harddisk for $200 per month. Just in case you need some arguments for HostGator

I am paying $130 for E3-1245, 32GB RAM and 2x 480GB SSD... and that is in Australia! Why are HG so expensive?

Because, reasons...

[gnif]

My hosting provider, 1and1, has full dedicated servers with 32 GB RAM, 12 cores x 2.3 GHz and 2 TB RAID 1 harddisk for $200 per month. Just in case you need some arguments for HostGator

I am paying $130 for E3-1245, 32GB RAM and 2x 480GB SSD... and that is in Australia! Why are HG so expensive?

Because, reasons...

I fully know why you use HG, and I support your position on this, just seems odd that they want so much for extra RAM. I have many client's hosted with WebNX in the US, dealings with WebNX have always been prompt, support has always been excellent, etc. We just upgraded a server from 64GB to 128GB of ECC RAM, and it cost an extra $40US per month. I get that HG have a custom hosting infrastructure and offering, the VM setup makes that very clear, but this setup should reduce their running costs, not push them up.

[EEVblog]

If I was to move hosts again then I wouldn't choose a US based one. Simple reason, DMCA.

I'd like to use an Australian host, just because, but given that 95% of my audience is elsewhere in the world, I figure that's not the best option.

[gnif]

If I was to move hosts again then I wouldn't choose a US based one. Simple reason, DMCA.

Ah yeah, I forgot about that fiasco. The issue is that most visitors to this website are US based, the performance is crazy for them with this new setup. It is great for us too, but we see 200-300ms of latency on the initial connection. If you were to go elsewhere I would suggest AU simply to make management a better experience, but I am biased from this point of view. I certainly would not consider any of the EU hosting offerings, the latency to EU from AU is terrible, exceeding 350-400ms. Hosting in China we get decent latency, around 200ms, but support isn't always great.

You could always host in Canada, for the US the latency is still great, for us it is about the same, but I don't know if the DMCA applies there.

[Fsck]

Have you considered OVH?
Their pricing is pretty competitive. (if you're in Canada, their pricing actually crushes everyone else due to the exchange rate now.)

[gnif]

Have you considered OVH?
Their pricing is pretty competitive. (if you're in Canada, their pricing actually crushes everyone else due to the exchange rate now.)

I have used OVH in Canada for quite some time, pros are costing and the scale of their network, etc... Cons, support is terrible, takes days often weeks to get an issue resolved. Most of their staff doesn't speak English, so the website auto-translates your tickets to French, which loses a lot. And you get what you pay for, they do not allow you to alter a host after the fact (ie, add RAM or a HDD).

They have just started offering hosting in AU which I have a test server with, so far this seems much better support wise then the CA support, but I suspect this is because the service is still considered experimental.

[Fsck]

you do get what you for with OVH though. (cheapest product = cheapest support, since you can't counterfeit hardware)

they're leagues cheaper than anyone else though.
if you want to customize and evolve over time, colo is really much better.

I don't have a lot of experience with hosting, but the dozen or so odd providers I've used that do allow upgrades charge silly monthly price increases for hardware upgrades. If all you want is data transfer, they'll do that for a good price, no problemo but want a single 4GB stick of ram? - that'll be 4$ a month + 30$ installation fee with no option to just outright buy one for them.

[gnif]

if you want to customize and evolve over time, colo is really much better.

The problem with collocation is that you now have to support the hardware and have spares on hand, possibly even rent some storage. If you need hands on you pay $100p/h on average. Since the server would not be local, Dave would have to pay for hands on, shipping parts, etc. In the end it is best to keep it simple and rent it. $4 + $30 install is extremely cheap compared with what HG are charging/gouging for.

Also if you had to pay for hands on, that's $100 + say $80-100 for the RAM, lets say $200. Where renting it at $4p/m, would take 50 months (4.16 years) to cost the same (not inc the install fee). Colo only makes sense if you are physically close to the DC to support your gear.

[cw360]

The head of marketing at HostGator (who is a fan BTW)

Gentlemen,

That guy is me..... At this point I should probably put on the body armor.... Outing yourself as a HostGator employee is typically seen as a risky move.

Dave is correct, I am a fan. I have been watching for years. I am a Marketing guy who plays with electronics on the weekends (arduino, raspberry pi, ham radio, 3d printing). But I should emphasize, much more Marketing than Engineering. The show and the forum have really helped me out with many personal projects. So a much delayed thank you to all of you.

Anyway, I am willing to field questions if you have them. If you throw anything too technical at me, it may take me a bit to find the answer or make sure it’s something we can actually share publicly. 

[gnif]

The head of marketing at HostGator (who is a fan BTW)

Gentlemen,

That guy is me..... At this point I should probably put on the body armor.... Outing yourself as a HostGator employee is typically seen as a risky move.

Dave is correct, I am a fan. I have been watching for years. I am a Marketing guy who plays with electronics on the weekends (arduino, raspberry pi, ham radio, 3d printing). But I should emphasize, much more Marketing than Engineering. The show and the forum have really helped me out with many personal projects. So a much delayed thank you to all of you.

Anyway, I am willing to field questions if you have them. If you throw anything too technical at me, it may take me a bit to find the answer or make sure it’s something we can actually share publicly.

Hi, nice to see you on here

I can say personally that I get a ton of use out of the website, videos and forum like yourself, and it is what drew me to offer my services as a way to give back. The questions I would like to ask I do not feel is my right to do so, as they are cost related and I will not be so bold to presume to think that Dave would like to see it discussed openly here.

From a technical POV we are pretty well informed and up to speed on how the HG hosting is setup, and I have spent many many hours tuning the server to run on your systems as optimal as possible. Our latest round of updates, both server side software changes and the memory increase have made an enormous difference. So I will make the same offer in return, if HG are interested in how we have this forum responding so well on this setup, please ask away, but note that some details are trade secrets.

[cw360]

The head of marketing at HostGator (who is a fan BTW)

Gentlemen,

That guy is me..... At this point I should probably put on the body armor.... Outing yourself as a HostGator employee is typically seen as a risky move.

Dave is correct, I am a fan. I have been watching for years. I am a Marketing guy who plays with electronics on the weekends (arduino, raspberry pi, ham radio, 3d printing). But I should emphasize, much more Marketing than Engineering. The show and the forum have really helped me out with many personal projects. So a much delayed thank you to all of you.

Anyway, I am willing to field questions if you have them. If you throw anything too technical at me, it may take me a bit to find the answer or make sure it’s something we can actually share publicly.

Hi, nice to see you on here

I can say personally that I get a ton of use out of the website, videos and forum like yourself, and it is what drew me to offer my services as a way to give back. The questions I would like to ask I do not feel is my right to do so, as they are cost related and I will not be so bold to presume to think that Dave would like to see it discussed openly here.

From a technical POV we are pretty well informed and up to speed on how the HG hosting is setup, and I have spent many many hours tuning the server to run on your systems as optimal as possible. Our latest round of updates, both server side software changes and the memory increase have made an enormous difference. So I will make the same offer in return, if HG are interested in how we have this forum responding so well on this setup, please ask away, but note that some details are trade secrets.

Whew, so far so good. Don't have to go into hiding yet.....

Glad you guys are here for Dave. I have been surprised when large youtubers/social media people (even corporate backed) have signed up for basic hosting (without talking to us) and then get surprised when their audience crushes a shared box. Speaks volumes about this community and Dave's foresight. Granted he is a born engineer turned social media guy, not original social media.

Re: your question: we took care of it.

Just throwing this out there. If you are ever interested in writing up a list of pro-tips, we are always looking for great guest posts on our blog. Would love to do a case study. Great way of getting some SEO love for the community. Not that this forum needs it in the slightest.

BTW, this thing is wicked fast.  I love/hate it when customers are faster than we are. lol

Another thing... Any thought to putting an SSL on here? We could probably make that happen......

[mrpackethead]

@cw360,

Do you have ipv6 at your site..

[cw360]

@mrpackethead

Hopefully I did that mention correctly. Had to go talk to the network guys about this one. They are actively educating me on data center architecture. Unfortunately, the official/short answer is: currently no.

[EEVblog]

The head of marketing at HostGator (who is a fan BTW)

Gentlemen,
That guy is me..... At this point I should probably put on the body armor.... Outing yourself as a HostGator employee is typically seen as a risky move.
Dave is correct, I am a fan. I have been watching for years. I am a Marketing guy who plays with electronics on the weekends (arduino, raspberry pi, ham radio, 3d printing). But I should emphasize, much more Marketing than Engineering. The show and the forum have really helped me out with many personal projects. So a much delayed thank you to all of you.
Anyway, I am willing to field questions if you have them. If you throw anything too technical at me, it may take me a bit to find the answer or make sure it’s something we can actually share publicly.

Big thanks.
FYI to all, HostGator came to the party and massively dropped the rates for the server with the new RAM 
Given how well the new server is now running, and the new low pricing, I'll be sticking with Hostgator.

[EEVblog]

Another thing... Any thought to putting an SSL on here? We could probably make that happen......

That's been discussed many times on here.
I'm sure gnif can fill you in the tech details as he has investigated this extensively I believe.
I would like to, as a lot of people have asked for it, but I believe there are some technical hurdles to overcome.

[gnif]

Another thing... Any thought to putting an SSL on here? We could probably make that happen......

That's been discussed many times on here.
I'm sure gnif can fill you in the tech details as he has investigated this extensively I believe.
I would like to, as a lot of people have asked for it, but I believe there are some technical hurdles to overcome.

The technical hurdles are no longer much of an issue anymore, the concern was always with performance. We have three options here

1) use CloudFlare for SSL, but that means we must proxy through them.
2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.
3) purchase a SSL certificate

[FrankBuss]

2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.

Interesting service. It is not much of a problem for this forum, but such a service sounds like the perfect way for the NSA or other 3 letter agencies to get private SSL keys.

[gnif]

2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.

Interesting service. It is not much of a problem for this forum, but such a service sounds like the perfect way for the NSA or other 3 letter agencies to get private SSL keys.

The private key never leaves the server, its a proper CA just has an API to perform automated renewals.

[CatalinaWOW]

Late to the response, but the upgrade has really been noticeable here from a satellite connection.  A few bumps during the transition, but extremely responsive by comparison.  Thanks to all who worked on it, and a big thank you to HostGator for making it affordable for Dave.

[xrunner]

No error messages for days - Yay! 

[EEVblog]

Another thing... Any thought to putting an SSL on here? We could probably make that happen......

That's been discussed many times on here.
I'm sure gnif can fill you in the tech details as he has investigated this extensively I believe.
I would like to, as a lot of people have asked for it, but I believe there are some technical hurdles to overcome.

The technical hurdles are no longer much of an issue anymore, the concern was always with performance. We have three options here
1) use CloudFlare for SSL, but that means we must proxy through them.

So if we chose to do that and we disable Cloudflare then the website revokes back to http only?

2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.
3) purchase a SSL certificate

Which type do we need?
http://support.hostgator.com/articles/ssl-certificates/acquire-ssl/how-do-i-purchase-an-ssl-and-what-type-is-it

[gnif]

So if we chose to do that and we disable Cloudflare then the website revokes back to http only?

Only if we do not install a cert on the server.

Which type do we need?
http://support.hostgator.com/articles/ssl-certificates/acquire-ssl/how-do-i-purchase-an-ssl-and-what-type-is-it

For this website since there is no real sensitive information, the cheapest option. You do not need to use HG for this, any certificate authority is fine, personally I have used RapidSSL, and AlphaSSL, but I believe there are better/cheaper options today. To be completely honest though if it was me I would just go for the letsencrypt.org service, it is if anything more secure then a generic SSL cert from other providers as it renews every 3 months, and the setup is quite simple on the server.

[amspire]

2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.

Interesting service. It is not much of a problem for this forum, but such a service sounds like the perfect way for the NSA or other 3 letter agencies to get private SSL keys.

LetsEncrypt went through a long period of public security auditing before they offered their service, so it is one of the more secure SSL solutions. It is done properly.

There are many hosting companies that support LetsEncrypt, and they will automatically renew every 3 months - so all you have to do is request a free SSL cert in the CPanel, and you have a certificate that will automatically renew for as long as you want.

Most CPanel hosts now let you enter your own key, so I have 3 LetsEncrypt certs I generate every 3 months (single batch file to update all three), and then I manually add the new keys to tye Cpanels.

For companies, I have always just got a commercial one - RapidSSL is something like A$12 a year if you buy from the right reseller.

Digicert is used by the likes of Facebook, so it is not a bad choice for companies.

[gnif]

2) use letsencrypt.org for a free SSL cert, this involves a little bit of setup on the server to renew the cert before it expires as these have a 3 month expiry.

Interesting service. It is not much of a problem for this forum, but such a service sounds like the perfect way for the NSA or other 3 letter agencies to get private SSL keys.

LetsEncrypt went through a long period of public security auditing before they offered their service, so it is one of the more secure SSL solutions. It is done properly.

There are many hosting companies that support LetsEncrypt, and they will automatically renew every 3 months - so all you have to do is request a free SSL cert in the CPanel, and you have a certificate that will automatically renew for as long as you want.

Most CPanel hosts now let you enter your own key, so I have 3 LetsEncrypt certs I generate every 3 months (single batch file to update all three), and then I manually add the new keys to tye Cpanels.

For companies, I have always just got a commercial one - RapidSSL is something like A$12 a year if you buy from the right reseller.

Digicert is used by the likes of Facebook, so it is not a bad choice for companies.

cPanel is not an option here for the renewal process, this server while runs cPanel is not using it for hosting, it's too slow. This is not an issue though as I can setup some custom scripts to handle the renewal.

[EEVblog]

For this website since there is no real sensitive information, the cheapest option. You do not need to use HG for this, any certificate authority is fine, personally I have used RapidSSL, and AlphaSSL, but I believe there are better/cheaper options today. To be completely honest though if it was me I would just go for the letsencrypt.org service, it is if anything more secure then a generic SSL cert from other providers as it renews every 3 months, and the setup is quite simple on the server.

I like the idea of a new certificate every few months from a "why not" point of view.
How much work is involved in setting this up?

[gnif]

For this website since there is no real sensitive information, the cheapest option. You do not need to use HG for this, any certificate authority is fine, personally I have used RapidSSL, and AlphaSSL, but I believe there are better/cheaper options today. To be completely honest though if it was me I would just go for the letsencrypt.org service, it is if anything more secure then a generic SSL cert from other providers as it renews every 3 months, and the setup is quite simple on the server.

I like the idea of a new certificate every few months from a "why not" point of view.
How much work is involved in setting this up?

About 1/2 hour, nothing huge.

[EEVblog]

About 1/2 hour, nothing huge.

Make it so when you have some free time!
I can add the SSL certificate to Cloudflare I believe if that's required.

[gnif]

About 1/2 hour, nothing huge.

Make it so when you have some free time!
I can add the SSL certificate to Cloudflare I believe if that's required.

Great, I will later today. In the CF configuration you just tell it to use "Strict" mode, then it enforces the SSL cert your server produces is valid.

[EEVblog]

About 1/2 hour, nothing huge.

Make it so when you have some free time!
I can add the SSL certificate to Cloudflare I believe if that's required.

Great, I will later today. In the CF configuration you just tell it to use "Strict" mode, then it enforces the SSL cert your server produces is valid.

Great, thanks.
Should start a new thread for this to report issues and discuss. Many nerds may rejoice.

[amspire]

For this website since there is no real sensitive information, the cheapest option. You do not need to use HG for this, any certificate authority is fine, personally I have used RapidSSL, and AlphaSSL, but I believe there are better/cheaper options today. To be completely honest though if it was me I would just go for the letsencrypt.org service, it is if anything more secure then a generic SSL cert from other providers as it renews every 3 months, and the setup is quite simple on the server.

I like the idea of a new certificate every few months from a "why not" point of view.
How much work is involved in setting this up?

The 3 month expiry is a great idea. If someone hacks your site and gets the keys to your commercial 2+ year SSL certificate, they can pretend to be you. You can revoke the certificate, but it is up to the browsers to check to see if certificates are revoked or not. As far as I know, Chrome does not actually check for revoked certs. They have their own system that can accept some revoked certificates.

[rrinker]

 I'm pretty sure Chrome blocks if the cert is revoked just like other browsers. Just had a customer attempt to replace their SHA-1 certs with SHA-256 and didn't realize that once he requests the existing cert to be re-keyed it revokes the old version of it, and attempting to access the site using this cert with Chrome resulted in a revocation error until it was replaced. Prior to rekeying, the error in Chrome was the one that since it was SHA-1 it wasn't actually secure.

[jippie]

Seems the forum is no longer sending HTTP_REFERER, which broke my dynamic avatar 

Stopped working probably somewhere between:
31/Mar/2017:13:03:04 +0200
and
31/Mar/2017:14:24:58 +0200

[amspire]

I'm pretty sure Chrome blocks if the cert is revoked just like other browsers. Just had a customer attempt to replace their SHA-1 certs with SHA-256 and didn't realize that once he requests the existing cert to be re-keyed it revokes the old version of it, and attempting to access the site using this cert with Chrome resulted in a revocation error until it was replaced. Prior to rekeying, the error in Chrome was the one that since it was SHA-1 it wasn't actually secure.

Last I heard, Google had decided not to let browsers check if a cert was revoked, but instead Google runs its own in-house tracking of revoked certs, and somehow Chrome uses this Google service. They did have an optional setting to re-enable revoke cert checking in Chrome, but that seems to have disappeared. This solution often works but not all the time. Google's cert database may be much faster, but it is not necessarily comprehensive or up-to-date.

If Google has gone back to allowing Chrome to directly check for revoked certs, it would be interesting to know.

[EEVblog]

Big jump in the number of forum sessions:

[cw360]

Big jump in the number of forum sessions:

Wow. This new traffic from somewhere? Google? I know they like SSLs.....

What do the source reports say?

If it's direct/eevblog.com there might be a self refer issue in the analytics.


Sent from my iPhone using Tapatalk

[tautech]

Big jump in the number of forum sessions:

There's been some crazy #'s viewing GK's thread:
https://www.eevblog.com/forum/projects/oscilloscope-pong-for-1-or-2-players/

[gnif]

What is the date of the jump?
What is the cycle of the peaks/troughs in the new session level? Weekly?
What is the old/new session length on average?

Like i've said before I'm no network expert but is there a session reuse or maybe it is called persistent sessions. Has that changed?

This coincides with fixing the 502 errors, I suspect CF has been serving tons of cached requests to try to work around the issue, we saw this jump before we turned on SSL.

[dimkasta]

I have letsencrypt on a couple of small sites for testing. I have no issues with it yet. Through cpanel, or using Laravel's homestead/forge.
You might want to check Taylor Otwell's scripts if you don't want to write them yourself. I think they are available with homestead.
Also, keep an eye on their newsletter. They recently changed their services, and caused disruption on certificate updates for many sites that did not update their scripts in time.

About cloudflare, keep an eye on your webmaster tools too. you might have to fiddle with seo settings or sitemap creation.
Google deindexed all images when we switched to https on one of our sites (without any obvious seo problems)

[tautech]

Short period of forum outages that lasted only a minute or two @~12.20 NZ time.
522 error.

[dimkasta]

Same here

[gnif]

Short period of forum outages that lasted only a minute or two @~12.20 NZ time.
522 error.

Thanks,

522 is a CloudFlare error, it is when CF can not establish a connection to the server, there has been no server faults or reported network outages on our end, I would guess that there was a temporary down route somewhere between CF and the server.

[tautech]

Gotta say how good the forum's been lately....just brilliant. 

Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

[gnif]

Gotta say how good the forum's been lately....just brilliant. 

Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

Thanks mate, I am glad to hear it

[tautech]

Gotta say how good the forum's been lately....just brilliant. 

Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

Thanks mate, I am glad to hear it

Thought I'd mention it as we all have a bitch quick enough when things aren't going right but we're often a bit slow to commend your efforts after all is well. Been meaning to say something for a few days but you know how it gets sometimes..............

[EEVblog]

Gotta say how good the forum's been lately....just brilliant. 
Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

I originally did put my hand in my pocket, but then full credit to HostGator who really came to the party and are now offering me the improved server for like half what I was paying before 
And yes, as always a huge thanks to gnif who spends a lot of time tweaking and maintaing the server for free 

BTW, I did get a 502 yesterday for like 30 seconds.

[gnif]

Gotta say how good the forum's been lately....just brilliant. 
Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

I originally did put my hand in my pocket, but then full credit to HostGator who really came to the party and are now offering me the improved server for like half what I was paying before 
And yes, as always a huge thanks to gnif who spends a lot of time tweaking and maintaing the server for free 

BTW, I did get a 502 yesterday for like 30 seconds.

No problem mate, do you recall roughly what time it occurred at?

[EEVblog]

Gotta say how good the forum's been lately....just brilliant. 
Not an error at all for weeks, just perfect.
Thanks Dave for putting your hand in your pocket and gnif for the grafting and tweaks. 

I originally did put my hand in my pocket, but then full credit to HostGator who really came to the party and are now offering me the improved server for like half what I was paying before 
And yes, as always a huge thanks to gnif who spends a lot of time tweaking and maintaing the server for free 

BTW, I did get a 502 yesterday for like 30 seconds.

No problem mate, do you recall roughly what time it occurred at?

Yesterday afternoon some time.

[gnif]

Yesterday afternoon some time.

Strange, I only see two errors for any of your IPs over the last two weeks, both of which are some bug in Sendy (see below, heavily redacted for security).

error.log-20170420.gz:2017/04/19 15:07:21 [error] 24952#0: *59443517 FastCGI sent in stderr: "PHP message: PHP Warning:  Cannot modify header information - headers already sent by (output started at redacted) in redacted on line redacted" while reading upstream, client: redacted, server: redacted, request: "GET redacted HTTP/1.1", upstream: "fastcgi://unix:redacted", host: "redacted", referrer: "redacted"
error.log-20170421.gz:2017/04/20 19:32:25 [error] 24953#0: *62660078 FastCGI sent in stderr: "PHP message: PHP Warning:  Cannot modify header information - headers already sent by (output started at redacted) in redacted: eval()'d code on line redacted" while reading upstream, client: redacted, server: redacted, request: "GET redacted HTTP/1.1", upstream: "fastcgi://unix:redacted", host: "redacted", referrer: "redacted"

Both of which are just header warnings, neither of which would have thrown a 502 error.

[EEVblog]

Glitch in the matrix

[PA0PBZ]

Yes it is working fine now, no more endless waiting for page loads and no more load errors.
I noticed something weird though, probably a forum software glitch? It's the "Local time" in the profiles.
For instance at the moment Gnif's profile says "Local Time:30 April 2017, 05:08:36" and for Dave it says "Local Time:29 April 2017, 18:10:23"
I'm sure you guys are not that far apart 

[RGB255_0_0]

Seems Nusa read wraper's mind.

https://www.eevblog.com/forum/renewable-energy/use-240v-household-type-mcbs-in-low-voltage-circuit/msg0/?topicseen#new

[PA0PBZ]

There is a profile setting in "look and layout" to set a time offset to correct the displayed time to your local time.

Yes I know, and I guessed almost everyone sets this to have the post times correct, but if I check the local time about 3/4 is incorrect. Maybe they just don't care?

[gnif]

Hah, I had not even noticed, I just corrected it! I set it to +11 out of habit without even thinking that the server's timezone is AEST already. Servers I manage generally run on UTC.

[Brumby]

... and here was me thinking you were that far ahead of the game.

[Monkeh]

Hah, I had not even noticed, I just corrected it! I set it to +11 out of habit without even thinking that the server's timezone is AEST already. Servers I manage generally run on UTC.

Perhaps this one should run on UTC too. Although that might make a mess of DB log/journaling timestamps during the changover. But one isolated server probably doesn't matter in the grand scheme of things.

Does that mean the time is changed for DST? Does the overlapping timestamps when going back an hour not create a problem?

Timestamps will all actually be stored in Unix time. Everything is actually in UTC (..close enough, anyway) - it just applies an offset for display.

It does cause everyone not in the same timezone to need to change their timezone four times a year on the forum, however, as SMF is, well, Simple.

[gnif]

Hah, I had not even noticed, I just corrected it! I set it to +11 out of habit without even thinking that the server's timezone is AEST already. Servers I manage generally run on UTC.

Perhaps this one should run on UTC too. Although that might make a mess of DB log/journaling timestamps during the changover. But one isolated server probably doesn't matter in the grand scheme of things.

Does that mean the time is changed for DST? Does the overlapping timestamps when going back an hour not create a problem?

Timestamps will all actually be stored in Unix time. Everything is actually in UTC (..close enough, anyway) - it just applies an offset.

It does cause everyone not in the same timezone to need to change their timezone four times a year on the forum, however..

I wonder how hard it will be to patch SMF to just do it properly... when I get a little time I will investigate.

[Monkeh]

I wonder how hard it will be to patch SMF to just do it properly... when I get a little time I will investigate.

That'd be nice.

A little brief searching suggests they've never implemented it because they don't like 'too many options' (okay, there are a lot of timezones, but SMF is presently as dumb as my microwave..).

[gnif]

I wonder how hard it will be to patch SMF to just do it properly... when I get a little time I will investigate.

That'd be nice.

A little brief searching suggests they've never implemented it because they don't like 'too many options' (okay, there are a lot of timezones, but SMF is presently as dumb as my microwave..).

It should be simple, they have already implemented the time offset stuff, clearly they are using the 'date' function everywhere instead of the 'gmdate' function.

[tautech]

Running Win 7 and time linked to some online clock gives zero problems for NZ daylight saving times.
Once you've got your offset from the forum default sorted and set, it works without issue.... does here anyway.

[Monkeh]

I wonder how hard it will be to patch SMF to just do it properly... when I get a little time I will investigate.

That'd be nice.

A little brief searching suggests they've never implemented it because they don't like 'too many options' (okay, there are a lot of timezones, but SMF is presently as dumb as my microwave..).

It should be simple, they have already implemented the time offset stuff, clearly they are using the 'date' function everywhere instead of the 'gmdate' function.

The issue appears to be with presenting the full range of options - if you just do a drop-down menu there are 561 options (and 385 unique zones, lots of links) in the tz database. Even cut down it's a long list. Some sort of graphical picker helps there, but requires more work to implement.

Personally I'd be happy with a box and instructions to locate the correct zone.

Running Win 7 and time linked to some online clock gives zero problems for NZ daylight saving times.
Once you've got your offset from the forum default sorted and set, it works without issue.... does here anyway.

DST start/end dates around the world do not align. I get to change it four times a year. This is boring.

[tautech]

Running Win 7 and time linked to some online clock gives zero problems for NZ daylight saving times.
Once you've got your offset from the forum default sorted and set, it works without issue.... does here anyway.

DST start/end dates around the world do not align. I get to change it four times a year. This is boring.

Yeah I get that and used to get the same, that's why I mentioned linking of your PC clock to an online source AND having the correct timezone and country set on your PC too. (didn't mention that before.)

Never need to adjust mine for the forums I frequent, the PC just sorts it for me.

[Monkeh]

Running Win 7 and time linked to some online clock gives zero problems for NZ daylight saving times.
Once you've got your offset from the forum default sorted and set, it works without issue.... does here anyway.

DST start/end dates around the world do not align. I get to change it four times a year. This is boring.

Yeah I get that and used to get the same, that's why I mentioned linking of your PC clock to an online source AND having the correct timezone and country set on your PC too. (didn't mention that before.)

My clocks are accurate. That does not stop the apparent server time changing.

[tautech]

Running Win 7 and time linked to some online clock gives zero problems for NZ daylight saving times.
Once you've got your offset from the forum default sorted and set, it works without issue.... does here anyway.

DST start/end dates around the world do not align. I get to change it four times a year. This is boring.

Yeah I get that and used to get the same, that's why I mentioned linking of your PC clock to an online source AND having the correct timezone and country set on your PC too. (didn't mention that before.)

My clocks are accurate. That does not stop the apparent server time changing.

Gotcha now.
Not seen that, it would piss me off too.
Is it really a SMF issue ? Or something else ? 

[gnif]

Gotcha now.
Not seen that, it would piss me off too.
Is it really a SMF issue ? Or something else ? 

Yeah, SMF is reporting the time in the server's timezone, it wouldn't be an issue if the client base was primarily in one area, as most forum user bases are. PHP provides two methods of dealing with the time, 'date' and 'gmdate', where date applies the server's timezone to the date, and gmdate is GMT. This is very common with open source projects like SMF, most programmers hate dealing with timezones due to how crappy they are, things like DST make it an absolute PITA to get it right.

Until you actually start to code time stuff, you don't (well at least I didn't) realize how involved it is, especially when you start to have to handle leap seconds, thankfully the kernel/os/PHP hides most of this from the dev, but it's can still be confusing. I find it is best to store times as UTC timestamps, then apply any offsets required when presenting the information, that way it never matters what the server or application's timezone is configured for.

[Monkeh]

Time quickly gets very complex.

If they're indeed simply using the 'wrong' function to print the time, at least switching to gmdate would restore DST changes to the usual twice a year. Still like to see timezone selection, though.

E: Seems to me the quickest solution is just to go into the admin panel and set the timezone to UTC where it belongs. Pretty sure that was shot down by Dave last time I raised it as he doesn't want to have to set his offset for DST..

[gnif]

E: Seems to me the quickest solution is just to go into the admin panel and set the timezone to UTC where it belongs. Pretty sure that was shot down by Dave last time I raised it as he doesn't want to have to set his offset for DST..

The server's logging, monitoring and other services are all relying on dates that are stored in the local timezone, while it is a quick setting to change, it will affect much more across the server and is too late to change it. If it is done the server's time to a ton of things, including the stats and backup systems will appear to jump back in time 10 hours, which will mess things up in a large way.

[Monkeh]

E: Seems to me the quickest solution is just to go into the admin panel and set the timezone to UTC where it belongs. Pretty sure that was shot down by Dave last time I raised it as he doesn't want to have to set his offset for DST..

The server's logging, monitoring and other services are all relying on dates that are stored in the local timezone, while it is a quick setting to change, it will affect much more across the server and is too late to change it. If it is done the server's time to a ton of things, including the stats and backup systems will appear to jump back in time 10 hours, which will mess things up in a large way.

Not on the server - on the forum. There is, according to the source at least (it's been a while since I ran SMF), a timezone setting on the forum. This may be something you can adjust without affecting much.

[EEVblog]

The SMF timezone is set to Sydney, always has been.
Any changes to any of this I'll leave up to gnif.