EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: TheSteve on September 13, 2016, 05:46:16 am
-
I have a raspberry pi I model B which has run web/named/ftp/ssh and some chat(Lambdamoo) services that has been connected to the internet pretty much since the pi I was released. I have recently replaced it with a Pi III running the same services. However I really marvel at just how reliable the Pi I has been. It flaked once since the original install with an sd card write problem that a reboot fixed. It has been super stable though since that one problem. While it is no longer "serving" anything I think I'll keep it running to see how long it goes.
Here is the uptime as of today:
steve@vip-orig ~ $ uptime
22:41:25 up 751 days, 26 min, 1 user, load average: 0.00, 0.01, 0.05
btw, during this time I have had many power outages. Most are short and the UPS handles them but there have been 3 extended ones where a generator was used - didn't want to lose the uptime!
Previous to the raspberry pi's I used an i5 laptop running linux, before that it was Sun Netra T1's running Solaris - they ran solid for 10+ years having never once crashed with the best time between reboots of over a year.
-
I once had a BSD server running for over 10 yrs and then decided to shut it down and add more ram to it, and clean it, well it didn't like the cleaning, or the ram ( the ram was fine./ and no conflict ).
got it all cleaned up and new ram installed and no power, it wound up being the power supply didn't like getting shut down aft 10 yrs of running, get a new power supply ( a used on i had laying around, as a test. )
and still no power. well it turns out that the power supply was bad, and the MB had died. and since it was better just to replace a 10+ yr old comp then to fix it, I replaced the hole thing.
-
3.5 years for a Windows XP Pro based "server" (actually a cheap netbook) running web, database, mail, network gateway with 2 NICs. Gave it to a friend once I stopped using it about 3 years ago and it's still being used regularly.
-
However I really marvel at just how reliable the Pi I has been.
I was looking for something to feed four HDMI displays with monitoring data for work. Instead of a four channel video cards and such thought I'd go cheap and hook four RPi-s running the Chromium browser. It's been a pretty crappy experience. Our monitoring suite generates pretty decent HTML, minimal JS, yet those browsers can't stay up for more than a week. Then the RPi-s started dying one by one, no idea what's happening to them, only figured recreating the SD card from scratch fixes it. For the most demanding of the web pages tried one of those Remix boxes. It's marginally faster and a bit more reliable. They are all toys. Or shouldn't be used with GUI, or something...
-
One of my Pis gave constant problems, crashing every couple of months and needing the card reflashed. That was fixed with a new SD card (which came with the realisation that not all 4GB cards are the same size, get one a few bytes smaller and you can't just drop the image onto it).
The other has been rock solid, but has outlived two power supplies. Both times fine after replacing the supply. That one gets used as a local file server with public facing http and ssh. Nowadays it gets mainly used as a chat server (with unix "talk", old but functional).
-
Most high upimes seems to be for kit that doesn't do much.
However I've got two CentOS 5 boxes that have been up for over 5 years. They are cache servers and have served approximately 34.9 requests/second each and that's just average over that time. The pair can push 10,000 a second out. I'll let you do the maths :)
-
Here is a pict of me shutting down a box that had been up for 1576 days:
-
One of my Pis gave constant problems, crashing every couple of months and needing the card reflashed. That was fixed with a new SD card (which came with the realisation that not all 4GB cards are the same size, get one a few bytes smaller and you can't just drop the image onto it).
The other has been rock solid, but has outlived two power supplies. Both times fine after replacing the supply. That one gets used as a local file server with public facing http and ssh. Nowadays it gets mainly used as a chat server (with unix "talk", old but functional).
And lots of SD cards are terrible for reliability, it's a crapshoot getting a good card if you're buying in general retail, I've been bitten a few times, the cards work fine for awhile but die at random intervals, a format and reload fixes it.
-
Damn, I thought mine was impressive.
I needed to check how long I had lived in this current place, so I checked my server lol. 174 days when I did.
I've had it up to a year I believe.
Do I get bonus points for it actually being a server though? :p
-
One of my Pis gave constant problems, crashing every couple of months and needing the card reflashed. That was fixed with a new SD card (which came with the realisation that not all 4GB cards are the same size, get one a few bytes smaller and you can't just drop the image onto it).
If you're using whatever crappy leftovers any 4GB card nowadays is, no wonder.
Get quality cards and you won't have an issue anymore.
-
I have similar. My Pi2 was up until I swapped it out for a 3. The 3 was up from whenever they were first available early this year until just last week when I upgraded from Slackware 14.2 to Slackware-current. This one runs owncloud and is my internet facing server.
-
Back in the 90's we had a printer server running Win NT3.51 with an uptime of a bit more than 4 years on a non-server class pc.
-
I have similar. My Pi2 was up until I swapped it out for a 3. The 3 was up from whenever they were first available early this year until just last week when I upgraded from Slackware 14.2 to Slackware-current. This one runs owncloud and is my internet facing server.
Slackware :-+ :-+ :-+
-
3 years uptime for a pair of Solaris based nameservers running in a carrier environment. Hardened by myself, direct internet access and no successful break-in. And they were an interesting target (tons of customer domains, several high profile ones).
-
the power supply didn't like getting shut down aft 10 yrs of running
I don't get this uptime stuff,
To me this justifies a policy of frequently rebooting every server as part of scheduled maintence.
I'd rather know in advance of a component on its last legs, than be surprised by it when I'm busy with other problems.
-
the power supply didn't like getting shut down aft 10 yrs of running
I don't get this uptime stuff,
To me this justifies a policy of frequently rebooting every server as part of scheduled maintence.
I'd rather know in advance of a component on its last legs, than be surprised by it when I'm busy with other problems.
It's usually not the reboot that kills it, but disconnecting the power. Same thing happened to my Pentium 166MHz router PC. Was working fine for years, also regularly updated the software (M0n0wall), until one day I had to move it. Pulled the cord, moved the machine, plugged it in and the PSU just blew up. Not sure what component failed, but the ceramic fuse inside was completely obliterated.
-
the power supply didn't like getting shut down aft 10 yrs of running
I don't get this uptime stuff,
To me this justifies a policy of frequently rebooting every server as part of scheduled maintence.
I'd rather know in advance of a component on its last legs, than be surprised by it when I'm busy with other problems.
I agree - in a commercial setting it is nice to know the machines will boot back up again and all services will start properly. In this case I was simply pleased a 35 dollar computer with a free OS is so stable.
-
I remember years ago when I worked for a Telco, we had a satellite office in Meadowbank (near Sydney) where a lot of our field techs worked out of. It was a very interesting building with many locked doors and unused rooms. The building used to be the office for Automatic Totalisators Limited, an Australian company that made ticket printing and totalisator machines. Some of the interesting aspects of the building included a locked safe room upstairs (we never did find the key to that) and under the reception area there were two trap doors which led to large "rooms" under the floor which weren't quite high enough to stand up in.
Anyway, getting back to the point, one day I was exploring the building because I had little to do. The receptionist had a stack of keys, so I just started trying doors. I found a room upstairs which I suppose was a utility/small store room. If I recall correctly there were a couple of racks of network equipment and a base station for a paging service (I guess they were just renting room in the building to fill a hole in their coverage).
There was also an old computer not connected to the network but still powered on. It was an old NT4 system (can't remember now if it was Server or Workstation), with some stupid rudimentary admin password like "password" or "123456". Anyway, it had an up time of several years.
-
So I don't have any uptimes to brag about, but talking about Raspberry Pi's brings up a couple things.
I read all over the net about how often memory cards are corrupted if you don't do a proper shutdown on a Pi before pulling power. I must be the luckiest Pi user then. I have a flaky USB cable powering my Pi and often if I move it or bump it, it just reboots. I must have done this a dozen times over the last few months. I've never corrupted the memory card. Yeah, I need a new cable so that doesn't happen, but it seems that all the rest of my micro USB cables are made with hair thin power wires that drop too much voltage under load and I get the rainbow power warning square in the corner of the screen when the CPU starts chugging hard. I need to go shopping for some good USB cables.
I did have a card go bad though, just a couple days ago. When Dave did the video on setting up the Pi with BOINC to crunch SETI units, I figured I might as well do that. I don't have my Pi2 doing anything useful right now anyway. I went in the other day and found it crashed. Figured the card finally got corrupted somehow and I'd have to reimage it. Problem is that when I try to reformat the card the OS says it's write protected. I've had this happen before on cards or flash drives that were written to heavily for a long time. I think the controller detects a problem or it exceeds a write count and it goes into read only mode to protect the files that are there. I figure BOINC did too much writing data to the card and wore it out.
And I thought these were good quality cards. Samsung EVO high speed cards...
Bought a new card, relearned how to properly format and image it, and the system is back up again. But I didn't put it back to work for NASA this time.
I've wondered how to set up the Pi 2 to do as little as possible with the memory card to boot up linux and run the rest of the OS from a small portable hard drive connected by USB. I have an old 30GB USB hard drive from a dozen years ago that would be plenty to run the Pi and not worry about wearing out a card.
It makes me feel old when I run down to Best Buy and find the 16GB cards on the clearance rack for $8, meaning when they are gone they won't even sell micro SD cards under 32GB.
-
I have issues getting uptime because they automatically sleep and wake up to save power and i update them in software and hardware often.
-
Currently the system with the longest uptime just hit the 1460 days today, or 4 years. A little celebration maybe?