Sunstone consumes memory

Opennebula 5.2.1

VM with frontend has 2Gb total memory

After each F5 (refresh) at Dashboard
/usr/bin/ruby /usr/lib/one/sunstone/sunstone-server.rb
consumes ~1% more memory up to >50% (1Gb)

Is it normal?

Denis Ldv opennebula@discoursemail.com writes:

Opennebula 5.2.1

VM with frontend has 2Gb total memory

After each F5 (refresh) at Dashboard
/usr/bin/ruby /usr/lib/one/sunstone/sunstone-server.rb
consumes ~1% more memory up to >50% (1Gb)

Is it normal?

I remember asking this question on IRC, I finished by setting up
Sunstone with Unicorn and each worker use several hundred MB of
non-swapped memory:

USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
oneadmin 23701  0.0  0.2  69244 13636 ?        Ssl  juin12   0:02 unicorn master -I /usr/lib/one/sunstone -c /etc/one/sunstone-unicorn.rb /usr/lib/one/sunstone/config.ru
oneadmin 23705  0.0  4.6 506412 282312 ?       Sl   juin12   0:34  \_ unicorn worker[0] -I /usr/lib/one/sunstone -c /etc/one/sunstone-unicorn.rb /usr/lib/one/sunstone/config.ru
oneadmin 23708  0.0  3.8 454020 233244 ?       Sl   juin12   0:36  \_ unicorn worker[1] -I /usr/lib/one/sunstone -c /etc/one/sunstone-unicorn.rb /usr/lib/one/sunstone/config.ru
oneadmin 23711  0.0  8.5 762124 520472 ?       Sl   juin12   0:46  \_ unicorn worker[2] -I /usr/lib/one/sunstone -c /etc/one/sunstone-unicorn.rb /usr/lib/one/sunstone/config.ru
oneadmin 23714  0.0  4.4 498352 274080 ?       Sl   juin12   0:47  \_ unicorn worker[3] -I /usr/lib/one/sunstone -c /etc/one/sunstone-unicorn.rb /usr/lib/one/sunstone/config.ru

This can result with my frontend swapping :-/

As I setup a memcached I can quite easily restart the unicorn service
without much trouble for my users since it’s quick and they are not
disconnected.

Regards.

1 Like

Thank you for answer.

I installed memcached and unicorn as you suggested. Plus https://github.com/kzk/unicorn-worker-killer for automatic worker restart after memory limit reach.

Denis Ldv opennebula@discoursemail.com writes:

Thank you for answer.

I installed memcached and unicorn as you suggested.

You are welcome. I said I should blog about the unicorn setup with
systemd services & co. but I never take the time to do it :-/

Plus https://github.com/kzk/unicorn-worker-killer for automatic
restart after memory limit reach.

Thanks, I was not aware of this one.

Regards.

I see this is a workaround but did anybody actually try to fix the memleak?
I have that on sunstone and on econe as well and while i don’t have a problem restarting sunstone i have problems restarting econe as i have constant load on it.

Had it on 5.2 and just added more memory but on 5.4 in ha this gets a bit excessive as i have 3 nodes where i have to increase memory.