On 08/04/2019 06.33, Andrei Borzenkov wrote:
08.04.2019 5:12, Carlos E. R. пишет: ...
Maybe related to this, I just saw something in the log. The swap in use is now above 6 GB.
Apparently the kernel is having some memory problem but coping, not crashing.
...
<0.4> 2019-04-08 01:06:41 Telcontar kernel - - - [335302.654173] Mem-Info: <0.4> 2019-04-08 01:06:41 Telcontar kernel - - - [335302.654177] active_anon:1159547 inactive_anon:197953 isolated_anon:0
You have 4.5GB of allocated memory that applications accessed recently. There is really nothing that can be done here short of adding more RAM or not using these applications.
<0.4> 2019-04-08 01:06:41 Telcontar kernel - - - [335302.654177] active_file:34229 inactive_file:19251 isolated_file:0 <0.4> 2019-04-08 01:06:41 Telcontar kernel - - - [335302.654177] unevictable:58 dirty:67 writeback:0 unstable:0 <0.4> 2019-04-08 01:06:41 Telcontar kernel - - - [335302.654177] slab_reclaimable:466310 slab_unreclaimable:37407
That's over 2GB of kernel memory. That looks too much; you may want to ask on LKML for ideas. There is slabtop command (do not know if it is installed by default) that shows owners of slabs and also how much is active; that may ring the bell for someone with more intimate kernel knowledge.
You may want to setup monitoring of memory consumption to see how it evolves.
Meanwhile, I updated and rebooted the machine. Just before that, hibernate aborted and was impossible. But no crash. I don't even know if those log entries indicate a problem, or what kind of problem (I was working at the machine most of the time and noticed nothing). I only wondered if this could be important/interesting enough to log a bugzilla here for someone to have a look. (yes, I know the machine would love more RAM, but it is maxed. Thus I would need a new machine, but I'm not prepared to spend that money now) -- Cheers / Saludos, Carlos E. R. (from 15.0 x86_64 at Telcontar)