Re: [Sks-devel] Extreme memory usage

[Todd Fleisher]

To Gabor’s point, I see some similar behavior on my nodes intermittently: https://i.imgur.com/aEePr6J.jpg

It seems to clear up on its own get automatically restarted by systemd after OOM kills it so for the most part it doesn’t appear to impact my pool from serving requests from the outside, presumably thanks to load balancing and/or caching. Also, it does not happen to all nodes at the same time.

FWIW - my nodes do not have the dump directory so that shouldn’t be at play here.

-T

On Jul 17, 2019, at 6:14 PM, Skip Carter <address@hidden> wrote:

Signed PGP part

If you do an 'lsof' you will also see all the dump files (223 when I
built it) are open, even when having done a normal build.  The only way
to prevent this is to move the dump directory after the build finishes.
This could also be contributing to the resource consumption.

On Wed, 2019-07-17 at 19:47 +0200, Kiss Gabor (Bitman) wrote:

Dear folks,

I experienced in the past months that my SKS instance dies almost
every day.
Now I think I found the reason.

Program crashes when e-mail address to be search is broken
into mostly 1-3 letter words.
These very short words - that are legitim search patters - make
sks db process eat the whole 6 GiB RAM of the VM and the OOM killer
reapes it.
Is this normal? What are your experiences?

Cheers

Gabor

_______________________________________________
Sks-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/sks-devel

--
Dr Everett (Skip) Carter
address@hidden

Taygeta Scientific Inc
607 Charles Ave
Seaside CA 93955
831-641-0645 x103

signature.asc
Description: Message signed with OpenPGP