I'm having problems with resource capping. In particular, a zone is able to
take out the whole machine which is clearly not what we want.
I have a zone capped at 4096M (of 8192) and is building mesos with make -j8
(it's an 8 core machine). As it gets into the 'meat' of the build, the rss
shoots up and locks the box - I think it must be thrashing pages.
I'm running the rcap daemon:
[root@i7 ~]# rcapadm
state: enabled
memory cap enforcement threshold: 75%
process scan rate (sec): 5
reconfiguration rate (sec): 60
report rate (sec): 5
RSS sampling rate (sec): 5
and it is enabled:
[root@i7 ~]# svcs | grep rcap
online 23:07:43 svc:/system/rcap:default
and just before the crash zonememstat reported
[root@i7 ~]# zonememstat
ZONE RSS(MB) CAP(MB) NOVER POUT(MB)
global 98 - - -
7302f2d4-a0a4-4686-9497-b9be94bf16d6 6159 4096 0 0
So maybe 75% is too high, but I think for a zone to be even able to take out
global is bad, bad news. Any ideas?
Thanks,
Dave
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
