Bug 15713 - Lease break unacked check triggered by network congestion leads to disconnect
Summary: Lease break unacked check triggered by network congestion leads to disconnect
Status: ASSIGNED
Alias: None
Product: Samba 4.1 and newer
Classification: Unclassified
Component: File services (show other bugs)
Version: unspecified
Hardware: All All
: P5 normal (vote)
Target Milestone: ---
Assignee: Ralph Böhme
QA Contact: Samba QA Contact
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2024-09-11 13:21 UTC by Ralph Böhme
Modified: 2024-09-12 08:07 UTC (History)
2 users (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Ralph Böhme 2024-09-11 13:21:48 UTC 
The problem is triggered when we send a lease break while the send queue is congested, so the lease break packet is queued behind other packets.

As we've implemented a unacked lease break logic to detect channel failures for multichannel, this currently quite short timer fires (TCP RTO * 6, so 1.5 seconds in the case we've debugged) and disconnects the channel and also exits if it's the only channel.

We have several ideas how this should be addressed including:

- prioritize lease breaks in Samba's send queue

- increase timeout

- don't use a timeout if there's only one channel

- restart timer every time the client consumes data (my favorite one :) )

- make the timeout configurable

- more ... ?