Project

General

Profile

Bug #950

Heavy Disk Usage causes Core to stop responding.

Added by masshuu almost 12 years ago. Updated over 11 years ago.

Status:
Feedback
Priority:
Normal
Assignee:
-
Category:
Quassel Core
Target version:
-
Start date:
04/22/2010
Due date:
% Done:

0%

Estimated time:
Version:
0.5.0
OS:
Any

Description

I have a core set up on a VPS, and it works great until the Disk Load picks up, at which point the core's lats sore to 300-800ms to irc servers(normally at 1-20ms) and the client to core lat shoots up to 10000+ms (normally 20-25ms)
Unlike quassel, which completely stops, everything else works just fine, apart from a slight change in speed.

History

#1 Updated by johu almost 12 years ago

  • Target version deleted (0.5.2)

wrong target version, 0.5.2 released months ago

#2 Updated by johu over 11 years ago

  • Status changed from New to Feedback

please add debug trace or steps to reproduce, otherwise it will be closed

#3 Updated by masshuu over 11 years ago

I am going to guess the issue has to do with calls to the database being synchronous.
I used to run the core on a VPS, which would occasionally become IO starved. This caused PostgreSQL to halt in its tracks, halting the core.
Then the DB got big enough that i couldn't login due to PostgreSQL taking a few seconds to process requests(i opened psql and hand timed simple queries, which took 1-2 seconds)
Simple reproduce would be to limit PostgreSQL's IO to something that would slow it down.

Also available in: Atom PDF