Time |
S |
Nick |
Message |
00:02 |
|
wajasu |
i've come to the conclusion that you can't run 2 plack workers with a 1cpu VPS. |
00:02 |
|
wajasu |
you can run 1cpu and bump the plack worker count to 1 |
00:03 |
|
wajasu |
or run a 2cpu VPS with 2 plack workers, |
00:03 |
|
wajasu |
or run 2 VPS 1cpu servers with 1 appserver/memcached, and 1 database. |
00:05 |
|
wajasu |
but 1cpu VPS with 2GB ram will work with 1 plack worker. |
00:05 |
|
wajasu |
when i had 2 plack workers trying to run on a 1 cpu server, one stalled and died. thus opac connection timeout. |
00:06 |
|
wajasu |
things worked with staff search, but opac does so many more requests, it took too long. |
00:07 |
|
wajasu |
my box from 10 years ago probably had an AMD with 2 cpus, plus it was koha version 3.8. |
02:33 |
|
tuxayo |
> i've come to the conclusion that you can't run 2 plack workers with a 1cpu VPS. |
02:33 |
|
tuxayo |
With 1 CPU, no concurrency is possible so having two workers shouldn't yield any gains while carrying a RAM and CPU overhead |
02:34 |
|
tuxayo |
> one stalled and died. thus opac connection timeout. |
02:34 |
|
tuxayo |
This shouldn't happen though ^^" |
02:35 |
|
tuxayo |
It should feel a bit slower if anything. (the CPU overhead shouldn't be much perceptible) |
02:36 |
|
tuxayo |
Anyway, does Koha still runs decently on your VPS with one CPU and one worker? |
02:36 |
|
wajasu |
it might be that 81 queries to the db is pushing it past a plack worker timeout. context switching? |
02:36 |
|
wajasu |
yes it runs as well as my old system. |
02:36 |
|
tuxayo |
good ^^ |
02:37 |
|
tuxayo |
> pushing it past a plack worker timeout |
02:37 |
|
tuxayo |
The timeout should take some time to happen tough. How fast it failed? |
02:38 |
|
wajasu |
4-5 seconds |
02:38 |
|
wajasu |
firefox browser is 90 by default. |
02:40 |
|
wajasu |
the other day, i had some local covers. so it was doing db queries, after memcached, plack worker contention, and serving 5 local covers. |
02:40 |
|
tuxayo |
> context switching? |
02:40 |
|
tuxayo |
Yes, it should happen a lot with 1 CPU for 1 worker and the DB. And even more for 2 workers (assuming multiple concurrent requests) but I don't know how much performance is really eaten by that. |
02:40 |
|
tuxayo |
Maybe there isn't much context switching with 1 worker. Since Koha waits for the DB call to end. |
02:40 |
|
wajasu |
i saw opac-image.pl in my top |
02:41 |
|
wajasu |
its faster with plack and memcached, than without. |
02:41 |
|
tuxayo |
> 4-5 seconds |
02:41 |
|
tuxayo |
That's way too fast. So something is wrong that it (webserver, worker or DB) timeouts that fast. |
02:41 |
|
tuxayo |
> its faster with plack and memcached, than without. |
02:41 |
|
tuxayo |
great :D |
02:42 |
|
wajasu |
if my librarian complains, i will bump to a 2cpu VPS for $30/month. i probably could do 2 $5 servers, with the db on one. and then backup to the second disk. |
02:43 |
|
wajasu |
i'm trying to find where I can set how many days back to keep the backups. |
03:02 |
|
|
bag joined #koha |
03:08 |
|
wajasu |
even though i migrated from 3.10 to 22.05, my database has all the notices with << >> template brackets. do all those notices need to migrate to [% %] or is the template toolkit configured to handle both. |
03:08 |
|
wajasu |
if it doesn't hadnle << >>, is there a script to migrate to [% %] ? |
04:46 |
|
|
ribasushi joined #koha |
07:26 |
|
|
cait joined #koha |
09:13 |
|
|
dpk_ joined #koha |
11:25 |
|
ashimema |
wajasu we still support both syntaxes for now.. not even all the default notices are migrated yet |
11:26 |
|
ashimema |
They support slightly different options though, some of the newer notices rely on TT feature not available in our custom <<>> replacements wtuff |
14:33 |
|
tuxayo |
git fetch returns "Gitea: Internal error" |
14:33 |
|
tuxayo |
But it managed to fetch commit anyway ^^" |
23:50 |
|
|
tcohen_ joined #koha |
23:50 |
|
|
sonOfRa_ joined #koha |
23:53 |
|
|
lari joined #koha |
23:53 |
|
|
jo_ joined #koha |
23:54 |
|
|
irc1 joined #koha |
23:55 |
|
|
fredericdem joined #koha |