Codehosting fails to release old connections when use_forking_daemon: True

Bug #732496 reported by John A Meinel
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Launchpad itself
Won't Fix
High
Unassigned

Bug Description

This is a follow up from the earlier "leaked handles". We tried enabling it again, but it yet again saw the number of connections continue to increase. Many connections connected and were released successfully, but somehow we would get ones that did not cleanup.
I'm marking this as a new bug, because the actual file handle leaks were fixed.

note that before we try this rollout again, we want to have switched over to a multi-server system, using haproxy. That way we can try one server in a new mode, without disrupting connectivity.

Revision history for this message
John A Meinel (jameinel) wrote :

Related to bug #717345. Probably the same root cause, but the extra file leaks caused the system to degrade much faster.

Revision history for this message
Robert Collins (lifeless) wrote :

We've rolledback to the older service, and this has never been live, so this is only high not critical.

Changed in launchpad:
importance: Critical → High
tags: added: canonical-losa-lp performance
William Grant (wgrant)
Changed in launchpad:
assignee: John A Meinel (jameinel) → nobody
Revision history for this message
Colin Watson (cjwatson) wrote :

We're going to be removing the forking service soon, so won't be fixing this bug. It was a decent attempt, but it never quite made it to production, and since the addition of git hosting it's unlikely that we'll be putting the resources into figuring out this sort of bzr optimisation project.

Changed in launchpad:
status: Triaged → Won't Fix
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.