can plugins+old ZFS cause instability?

[petr]

I've got a standard build, which I've been using without any stability issues with the 8.3.1. Then, I've done two things - upgraded to the latest FreeNAS (9.1.0 and later to 9.1.1) and started playing around with plugins and jails. I've not yet upgraded my pools to the latest versions.

I totally love the idea and went on and set up couchpotato, sickbeard, sabnzbd and few others that I did not configure yet, all from the official repository.

From there, I've started having whole system lockups - it happens mainly when interacting with couchpotato, the jail becomes unresponsive, soon taking the rest of the system with it. When the lockup happens, no system messages are output on the screens and I cannot log in via SSH, although the UI works. The only message I was able to catch a glimpse of was a UPS error right before the lockup, possibly indicating that the USB subsystem has crashed? (just guessing, in deep waters here).

I'm was using on-board Realtek NIC, so I've added Intel-based additional NIC and tried again - will wait to see if the lock-up happens again.

Lastly, after a clean reboot, sabnzb and sickbeard do not start for around 10 minutes (and I do not know if they would start on their own as I never resisted to try out the "on" button.

So, back to the main question - could a jail cause whole system instability? I find it too much of a coincidence that my machine's hardware would start failing at the same time as I've started using jails.


My hardware config:
Motherboard:
GIGABYTE GA-Z68X-UD5-B3

RAM:
12GB of 1600 of Corsair RAM (2x2GB and 2x4GB

CPU:
Intel Core i5 2300 2.8GHz Socket 1155

PSU:
Coolermaster Silent Pro 1000W Modular PSU

UPS:
APC – BACK UPS ES 700G

 

[dlavigne]

A jail shouldn't, though an overloaded NIC could. Has the system locked up since replacing the NIC?

 

[petr]

A jail shouldn't, though an overloaded NIC could. Has the system locked up since replacing the NIC?

The Intel NIC seems to be holding things together now - uptime almost 2 days and counting (and I am running scrubs to increase load further). Will hold my breath to see if it gets to a week, then I will be a bit more calm. I've also unplugged all USB peripherals bar the flashdrive (but those were there before running happily for a few months on 8.3.1).

My guess is that there are some problems in the kernel with the realtek driver and jail's virtual ethernet driver?

 

[dlavigne]

Again, I wouldn't blame it on the jail. Realteks are notorious for bad performance, especially under load.

 

[petr]

Again, I wouldn't blame it on the jail. Realteks are notorious for bad performance, especially under load.

I know - but one thing is bad performance and another is lockup. My only reason why to take the jails into equation is that it is the only variable that changed - I've been using the realtek NIC for a few months before without a single lockup. Either way, I am not complaining for the need to add the extra NIC - very low price for stability if you ask me!

 

[cyberjock]

Realtek's can be good/bad for both performance and reliability. Some people get good performance with Realteks, but nobody really beats Intel for performance. When it comes to reliability lots of people have reliability problems with Realteks. Intel's on the other hand are usually failed hardware, user error(unplugged cable), or the driver isn't part of FreeNAS. There's plenty of threads of people that had weird issues with Realtek NICs that were nondescript in the cause, but as soon as they went to Intel the problems went away.

That's why my presentation says to just use Intel and be done with it, otherwise you get to play the lottery. It might work fine or it might not. It might perform fine, or it might not. And for the amount of time and effort you(and forum responders) will put into explaining the whole mess you could just take the advice and put an Intel in and be done with it. :)

I'd recommend you just push the "I believe" button and try an Intel NIC. They aren't expensive but may save you boatloads of time trying to find an issue that you might not be able to fix anyway assuming its the NIC. At the worst you have a spare high quality NIC that was cheap that you can use for troubleshooting in the future. At the best your problem is solved.

 

[petr]

Realtek's can be good/bad for both performance and reliability. Some people get good performance with Realteks, but nobody really beats Intel for performance. When it comes to reliability lots of people have reliability problems with Realteks. Intel's on the other hand are usually failed hardware, user error(unplugged cable), or the driver isn't part of FreeNAS. There's plenty of threads of people that had weird issues with Realtek NICs that were nondescript in the cause, but as soon as they went to Intel the problems went away.

That's why my presentation says to just use Intel and be done with it, otherwise you get to play the lottery. It might work fine or it might not. It might perform fine, or it might not. And for the amount of time and effort you(and forum responders) will put into explaining the whole mess you could just take the advice and put an Intel in and be done with it. :)

I'd recommend you just push the "I believe" button and try an Intel NIC. They aren't expensive but may save you boatloads of time trying to find an issue that you might not be able to fix anyway assuming its the NIC. At the worst you have a spare high quality NIC that was cheap that you can use for troubleshooting in the future. At the best your problem is solved.

Thank you! That's what I've done.. and to be on a safe side, also used a new flash drive with latest image+db export rather than using the old, upgraded version - although that should not make a difference in my opinion..

I've got a new problem now, will probably start a new topic. After a few days of uptime (using the intel NIC with onboard realtek disabled), ping over local network grows to around 300-500ms. All services still work but much slower. When I ssh into the machine, nothing seems to be using CPU excessively and everything works. Also the GUI works. When I reboot through the ssh, it all starts working fast again. I am on a third cycle now, will keep track how long it takes for things to slow down again.