Re: [Xen-devel] [xen-4.4-testing test] 52629: regressions - FAIL

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [Xen-devel] [xen-4.4-testing test] 52629: regressions 
- FAIL"):
> Going through other recent flight failures I'm seeing all of these a
> fair bit. Searching for this particular message led me to
> https://bugs.launchpad.net/ubuntu/+source/openssh/+bug/416264
> 
> That suggests "-o GSSAPIAuthentication=no" as a workaround.

I think this is a red herring.  Our reverse DNS is not set up, but we
do have empty zones for all the relevant reverse ranges, so reverse
DNS lookups will fail quickly.

As I said on IRC:

13:49 <Diziet> Yesterday I found a copy of `convert' using 3G of RAM and the 
               osstest VM had a load of 30
13:51 <Diziet> 
http://logs.test-lab.xenproject.org/osstest/logs/52629/test-amd64-amd64-pv/20.ts-repeat-test.log
               shows ssh taking >10s to get to the point of 
               printing the remote protocol version, and getting timed out 
               halfway through enumerating its local private keys.
13:51 <Diziet> I sent it a SIGSTOP.  Let me find it...
13:53 <Diziet> Looks like that particular convert started on the 30th but a 
               previous run of the same bisection attempt (for that is what it 
               was) could explain a fair few spurious timeouts.
13:53 <Diziet> I have a patch queued up to change cs-bisection-step to use pbm 
               instead.
13:54 <Diziet> (And also fix a performance problem with some of the sql in 
               sg-report-job-history)

I'm waiting for the C600/TG3 bugfix to make it into osstest production
(because that's holding up the deployment of 4-6 test hosts) and then
I will probably force push those.

I'll email those patches later today

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel