Hi,
Sorry for the little delay, I had to learn OOM killer and finally
migrated my trafficserver instance to FreeBSD 9.3 (a FreeBSD jail
running on FreeNAS 9.3) where I have now 12 GB physical RAM (and 18 GB
virtual RAM) with constantly 3 GB physical RAM available and no way to
figure out memory limitations in jail (`rctl` prints `rctl:
rctl_get_rules: Function not implemented` both in jail and on host),
assuming there're none.
The process gets killed approximately every 30 seconds/10 MB and
trafficserver behaves like on the Linux machine.
Any further help is appreciated.
Best regards,
Kalle
Am 31.01.2015 um 04:46 schrieb James Peach:
>
>> On Jan 30, 2015, at 7:36 PM, Karl-Philipp Richter <rich...@richtercloud.de>
>> wrote:
>>
>> Hi,
>> I'm experiencing systematic failures of transfers of large files (> 5
>> MB) due to
>>
>> [Jan 31 04:22:05.856] Manager {0x2ab59000} ERROR:
>> [LocalManager::pollMgmtProcessServer] Server Process terminated due to
>> Sig 9: Killed
>
> Something killed traffic_server with SIGKILL? Was that the kernel OOM killer?
>
>> [Jan 31 04:22:05.857] Manager {0x2ab59000} ERROR:
>> [Alarms::signalAlarm] Server Process was reset
>>
>> which I found in `manager.log`. I read
>> https://docs.trafficserver.apache.org/en/latest/sdk/troubleshooting-tips/unable-to-debug-tags.en.html
>> in order to get started with debugging and turned on debugging for
>> `cache.*`, but there're no log entries related to the manager (neither
>> in `manager.log` nor in `traffic.out`). Is there any way I could get
>> around running the `LocalManager::pollMgmtProcessServer` in `gdb`
>> (because my skills are rather low in that area) and figure out enough
>> background information to file a useful bug report with logging only?
>>
>> Any help and pointers are appreciated!
>>
>> Best regards,
>> Kalle
>
