[OpenIndiana-discuss] Server hangs weekly

Sun Mar 25 16:06:45 UTC 2012

Just to close the loop on this thread:

After doing some performance investigations, I didn't see anything to
suggest memory issues, but I did notice kernel CPU usage was slowly rising
after every boot. That led me to this bug:

https://www.illumos.org/issues/1333

which is fixed at least as of oi151a_prestable1. (Oddly the workaround
suggested in that thread wouldn't take on my system--not sure why.) Having
upgraded, things have been running smoothly for some time now, and kernel
CPU usage has remained in check.

Thanks for the help.

On Fri, Feb 24, 2012 at 5:45 AM, Milan Jurik <milan.jurik at xylab.cz> wrote:

> Hi,
>
>
> On 23.02.2012 06:37, oimltalk at skidde.net wrote:
>
>> Interesting. I have 8GB of memory, which I thought would be enough for my
>> purposes (6 x 2TB drives, RAIDZ-2, no deduping or anything special like
>> that). Most of the time the server is on it's idle, but perhaps the
>> regular
>> snapshots are causing problems. I'll have to try to track memory usage and
>> see what happens.
>>
>>
> ZFS ARC is very aggressive in memory consumption. Together with other
> system caches it can consume a lot of free space and it is very lazy to
> give it back to system. And there seem to be some very bad reactions of
> system to OOM.
>
>
>  As for NWAM, I believe I'm using it? I haven't turned it off, anyway.
>>
>>
> so it could explain why you loose network. It could be NWAM daemon itself
> or some bad behavior in network stack giving info to NWAM daemon that
> network card lost link and later it is unable to reconfigure card. Probably
> separate bug.
>
> Best regards,
>
> Milan
>
>  Thanks for the feedback.
>>
>> On Wed, Feb 22, 2012 at 5:27 PM, Daniel Kjar <dkjar at elmira.edu> wrote:
>>
>>  I have this problem on a system that I was using to back up 50 gbs of
>>> material each night.  It would transfer that across the network in zfs
>>> and
>>> that would kill it but it would only happen after a week or so of nightly
>>> updates of roughly the same size.  This machine has 32gb of ram and a cp
>>> process would hang and swallow it all bringing the system to its knees.
>>>  I
>>> just stopped that big transfer job and called it a night.  I am no longer
>>> backing up my files to 3 different buildings but that is better than
>>> crashing my sunray server every 5 days.
>>>
>>>
>>> On 2/22/2012 1:48 AM, Milan Jurik wrote:
>>>
>>>  Hi,
>>>>
>>>> one of my systems was suffering from very similar symptoms. I had no
>>>> chance to debug it much as it was on remote site in serverhouse. But in
>>>> my
>>>> case it was lack of memory, system was under significant memory
>>>> pressure. I
>>>> was unable to reproduce it on small systems I have at home. I added some
>>>> memory and set limits for zones.
>>>>
>>>> One small suggestion - could you write small script dumping memory info
>>>> (from kernel mdb) and list of processes to the disk and run it from
>>>> crontab
>>>> every few minutes? Maybe it will be unable to store data during "hang"
>>>> but
>>>> at least you could see trend.
>>>>
>>>> For lost IP address - are you using NWAM?
>>>>
>>>> Best regards,
>>>>
>>>> Milan
>>>>
>>>> On 22.02.2012 07:32, oimltalk at skidde.net wrote:
>>>>
>>>>  Hi there,
>>>>>
>>>>> I'm seeing roughly weekly hangs on a server running OpenIndiana 151a.
>>>>> I'm
>>>>> using it primarily as a home fileserver with ZFS.
>>>>>
>>>>> The exact behavior seems to depend on when I notice it, but essentially
>>>>> the
>>>>> server drops off the network and is only variably responsive when I try
>>>>> to
>>>>> access the console directly. Sometimes when this happens the system
>>>>> doesn't
>>>>> respond at all (e.g., not even to keyboard input). One time I was able
>>>>> to
>>>>> interact with the console (after the server had disappeared from the
>>>>> network) and tried to see what was going on. Tried pinging
>>>>> google.com(unreachable, as expected). Next I tried `ifconfig -a` and
>>>>> got this:
>>>>>
>>>>> lo0: flags=2001000849<UP,LOOPBACK,****RUNNING,MULTICAST,IPv4,****
>>>>> VIRTUAL>
>>>>>
>>>>> mtu 8232
>>>>> index 1
>>>>>       inet 127.0.0.1 netmask ff000000
>>>>> e1000g0: flags=1040843<UP,BROADCAST,****RUNNING,MULTICAST,DEPRECATED,*
>>>>> ***IPv4>
>>>>>
>>>>> mtu
>>>>> 1500 index 2
>>>>>       inet 0.0.0.0 netmask ff000000
>>>>>
>>>>>
>>>>> which explains the lack of connectivity. But after it printed that it
>>>>> didn't return. The console still printed my keyboard output (including
>>>>> ^C,
>>>>> ^Z, etc.), and there was still output coming from other sources (e.g.,
>>>>> I
>>>>> have napp-it running regular snapshots, so I saw a notice that it had
>>>>> used
>>>>> sudo to run that) but I couldn't get a prompt back. Next I tried
>>>>> hitting
>>>>> the power button on the machine I got this:
>>>>>
>>>>> poweroff: initiated by user on /dev/console
>>>>> in.ndpd[994]: phyint_reach_random: SIOCSLIFLNKINFO (interfac e1000g0):
>>>>> Interrupted system call
>>>>> bootadm: /boot/solaris/bin/extract_****boot_filelist is not owned by
>>>>> 101,
>>>>>
>>>>> skipping
>>>>> syncing file systems... done
>>>>> WARNING: Power off requested from power button or SC, powering down the
>>>>> system!
>>>>>
>>>>>
>>>>> followed shortly by:
>>>>>
>>>>> WARNING: Failed to shut down the system!
>>>>>
>>>>>
>>>>> Tried looking through the logs for anything interesting but didn't come
>>>>> up
>>>>> with anything, though to be honest I'm not 100% sure where to look or
>>>>> what
>>>>> to look for. When the machine drops off the network I can still access
>>>>> it
>>>>> via IPMI (tried this using both the dedicated jack on the motherboard
>>>>> and
>>>>> by sharing the Intel NIC--worked in both cases, but OI was still
>>>>> unresponsive), so I doubt it's a bad NIC. Motherboard is a Supermicro
>>>>> X9SCM-F.
>>>>>
>>>>> I know that at least sometimes the system will stop running even my ZFS
>>>>> snapshots via napp-it, since I've come back to a frozen console that
>>>>> showed
>>>>> the last snapshot being taken 12+ hours before (they're supposed to be
>>>>> taken every 15 minutes). My guess is this is just because it takes me
>>>>> longer to notice sometimes--seems like it's hitting a deadlock
>>>>> somewhere
>>>>> that eventually grinds everything to a halt (like with the ipconfig
>>>>> call
>>>>> above).
>>>>>
>>>>> Also, FWIW, here's what ipconfig -a gets me when it works correctly
>>>>> (MAC
>>>>> address removed, although interestingly it wasn't even printed in the
>>>>> output above):
>>>>>
>>>>> lo0: flags=2001000849<UP,LOOPBACK,****RUNNING,MULTICAST,IPv4,****
>>>>> VIRTUAL>
>>>>>
>>>>> mtu 8232
>>>>> index 1
>>>>>       inet 127.0.0.1 netmask ff000000
>>>>> e1000g0: flags=1040843<UP,BROADCAST,****RUNNING,MULTICAST,DHCP,IPv4>
>>>>> mtu
>>>>>
>>>>> 1500
>>>>> index 2
>>>>>       inet 192.168.10.10 netmask ffffff00
>>>>>       ether [MAC address here]
>>>>> lo0: flags=2001000849<UP,LOOPBACK,****RUNNING,MULTICAST,IPv6,****
>>>>> VIRTUAL>
>>>>>
>>>>> mtu 8252
>>>>> index 1
>>>>>       inet6 ::1/128
>>>>> e1000g0: flags=20002004841<UP,RUNNING,****MULTICAST,DHCP,IPv6> mtu
>>>>> 1500
>>>>>
>>>>> index 2
>>>>>       inet6 fe80::225:90ff:fe50:2c2a/10
>>>>>       ether [MAC address here]
>>>>>
>>>>>
>>>>> Any ideas/suggestions on where to go from here? Thanks in advance.
>>>>>
>>>>>
>>>>>
>>>>
> ______________________________**_________________
> OpenIndiana-discuss mailing list
> OpenIndiana-discuss@**openindiana.org<OpenIndiana-discuss at openindiana.org>
> http://openindiana.org/**mailman/listinfo/openindiana-**discuss<http://openindiana.org/mailman/listinfo/openindiana-discuss>
>