Description of problem:
If VDSM gets a stomp frame with the content-length header and the frame cuts off just missing the terminating '\0' the frame parser will have issues parsing subsequent messages.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
Barak, can you provide steps how to verify this?
This bug is time dependent and it is very hard to have a reproducer. When vdsm code handles messages there maybe an issue when network frame contains most of the message except of '\0'. When this rare situation occurs parser was not able to process the message.
We run a unit test with several thousand operations to find this issue.
Piotr, have you had any luck with reproducing the problem running your unit tests?
Or is there another way how QA can for example fake or tamper STOMP frames in order to verify the bug?
I do not see this issue in my tests anymore. You would need to create a client which generates heavy load on stomp server ~5k messages per second.
Piotr, can you please help us again to verify this bug, the same way as in 3.5.1 (bug 1195116 comment 2) ?
Can you run your unit tests against 3.6 VDSM if you manage to reproduce or not?
I run the tests for 3.6 and I do not see this issue.
(In reply to Piotr Kliczewski from comment #11)
> I run the tests for 3.6 and I do not see this issue.
Thank you, Piotr.
Based on this I am moving the bug to VERIFIED (vdsm-4.17.8-1.el7ev, build 3.6.0-15).
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.