Performance: Do not schedule Session.purge if this Session is not used

[isimluk]

You can use memcache OR SQL session. When you use memcache this whole job does nothing. Just inserts into a queue, updates queue, selects empty from Session and then removes from queue.

This also frees up some memory and scheduler cycles. Yaix!

Note, we already require users (in documentation) to restart EVM server when they change session store (so we can do this).

[miq-bot]

Checked commit isimluk@d393bef with ruby 2.2.6, rubocop 0.47.1, and haml-lint 0.20.0
2 files checked, 0 offenses detected
Everything looks fine. 👍

[martinpovolny]

@imtayadeway, @gtanzillo : any reason not to merge this one?

Makes sense and the code looks good to me.

[gtanzillo]

👍 I'm good with this change

[kbrock]

This has minimal to no performance impact - 1 query returning 0 records.
But in my environment, this represents 39% of the messages processed by the GenericWorker.

And, it is also only applicable to customers running with AR sessions (basically if they are running a load balancer)

I would like to backporting to fine/yes

[isimluk]

Sounds good.

Cheers!

🛀

[isimluk]

This has minimal to no performance impact - 1 query returning 0 records.

Actually to do this nothing job we need to:

• insert the job into queue
• select the job from the queue
• update the job in the queue (say its in progress)
• process -> do nothing -> select empty
• delete the job from the queue

And now ... let me count the syscalls...

[simaishi]

@kbrock Is there a BZ for this? Need a BZ to backport to Fine branch...

[kbrock]

Thanks @simaishi https://bugzilla.redhat.com/show_bug.cgi?id=1510040

[simaishi]

Fine backport details:

$ git log -1
commit 15889ffdf20cdd4f13a11c00a19b931f0808e13e
Author: Martin Povolny <[email protected]>
Date:   Tue May 16 15:39:19 2017 +0200

    Merge pull request #15064 from isimluk/avoid-queue-overhead-if-session-is-turned-off
    
    Performance: Do not schedule Session.purge if this Session is not used
    (cherry picked from commit 493a4884dfd141d7cb3640446d35738c67e42e69)
    
    https://bugzilla.redhat.com/show_bug.cgi?id=1510054