DirtyCleanInterval should be 0 by default

Bug #1830022 reported by Dariusz Gadomski
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
cups (Ubuntu)
Fix Released
Low
Till Kamppeter

Bug Description

Please consider changing DirtyCleanInterval value to 0 as default.

Otherwise if cupsd crashes due to (e.g. OOM killer) under a heavy workload even hundreds of jobs may be lost. This concern is backed up by a real-life scenario and leaves the client sending thousands of jobs unaware that many of them are lost during a crash.

After cupsd gets restarted it rewinds it's job counter to the last cached and continues unaware about the jobs accepted and lost.

Having DirtyCleanInterval set to 0 will cause some performance impact, but not significant under lighter workloads and a completely justified price for reliability under heavy workloads.

Test scenario:
1. sudo apt install printer-driver-cups-pdf
2. while [ 1 ]; do lp -d PDF somepdf.pdf; done;
3. # on other terminal
   kill -9 $(pidof cupsd)
4. Note last job number and wait for cupsd to be restarted by systemd.
5. Once accepting jobs is resumend the job counter is rewound.

Expected behavior:
Accepted jobs are queued for processing.

Actual behavior:
Some accepted jobs are lost.

Revision history for this message
Launchpad Janitor (janitor) wrote :

Status changed to 'Confirmed' because the bug affects multiple users.

Changed in cups (Ubuntu Bionic):
status: New → Confirmed
Changed in cups (Ubuntu Cosmic):
status: New → Confirmed
Changed in cups (Ubuntu Disco):
status: New → Confirmed
Changed in cups (Ubuntu Xenial):
status: New → Confirmed
Changed in cups (Ubuntu):
status: New → Confirmed
no longer affects: cups (Ubuntu Xenial)
no longer affects: cups (Ubuntu Bionic)
no longer affects: cups (Ubuntu Cosmic)
no longer affects: cups (Ubuntu Disco)
tags: added: rls-nn-incoming
tags: removed: rls-nn-incoming
Revision history for this message
Dariusz Gadomski (dgadomski) wrote :
tags: added: rls-ee-incoming
Changed in cups (Ubuntu):
assignee: nobody → Till Kamppeter (till-kamppeter)
Revision history for this message
Sebastien Bacher (seb128) wrote :

The problem there doesn't seem common enough to justify being rls tracked, would still be nice to be fixed and it looks like it's being worked without need of special handling, tagging rls-ee-notfixing

Changed in cups (Ubuntu):
importance: Undecided → Low
tags: added: rls-ee-notfixing
removed: rls-ee-incoming
tags: added: sts
Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package cups - 2.2.11-0ubuntu1

---------------
cups (2.2.11-0ubuntu1) eoan; urgency=medium

  * New upstream release
  * Removed 10 patches taken from upstream.
  * Refreshed 13 patches with quilt.
  * Removed patch to set "LogDebugHistory 99999" as it causes high
    system load when there are very many jobs (LP: #1831021).
  * Set "DirtyCleanInterval 0" in CUPS configuration to avoid losses of
    jobs when cupsd gets killed (LP: #1830022).

 -- Till Kamppeter <email address hidden> Mon, 05 Aug 2019 11:40:10 +0200

Changed in cups (Ubuntu):
status: Confirmed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.