teuthology-suite and priorities

teuthology-suite and priorities

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 852 bytes --]

Hi,

When scheduling suites that are low priority (giant for instance at http://pulpito.ceph.com/loic-2015-03-23_01:09:31-rados-giant---basic-multi/), the --priority 1000 is set because (if I remember correctly) this is the standard priority for nightlies. However, it seems to be different because this morning the giant suite was running although a hammer suite was scheduled before it (for instance http://pulpito.ceph.com/teuthology-2015-03-22_23:18:01-multimds-hammer-testing-basic-multi/) and was still waiting.

The giant suite was killed so that it does not interfere with more urgent jobs. Looking at the teuthology jobs there does not seem to be a trace of the priority, probably stored elsewhere. 

I would be very grateful if you could explain how priorities should be used.

Cheers

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: teuthology-suite and priorities

[Gregory Farnum]

On Mon, Mar 23, 2015 at 7:20 AM, Loic Dachary <loic@dachary.org> wrote:
> Hi,
>
> When scheduling suites that are low priority (giant for instance at http://pulpito.ceph.com/loic-2015-03-23_01:09:31-rados-giant---basic-multi/), the --priority 1000 is set because (if I remember correctly) this is the standard priority for nightlies. However, it seems to be different because this morning the giant suite was running although a hammer suite was scheduled before it (for instance http://pulpito.ceph.com/teuthology-2015-03-22_23:18:01-multimds-hammer-testing-basic-multi/) and was still waiting.
>
> The giant suite was killed so that it does not interfere with more urgent jobs. Looking at the teuthology jobs there does not seem to be a trace of the priority, probably stored elsewhere.
>
> I would be very grateful if you could explain how priorities should be used.

I've seen a couple of oddities with suites that you've scheduled
before, and I think something in the system might not be using
timestamps properly. :) Also, the multimds suite is not an important
one right now, so feel free to zap it rather than an LTS point release
run in future.

1000 is definitely the default, and I don't have one handy to look at
right now but I think if you set a non-default priority it will show
up in one of the config files.
-Greg

Re: teuthology-suite and priorities

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 1881 bytes --]


Hi Greg,

On 23/03/2015 15:24, Gregory Farnum wrote:
> On Mon, Mar 23, 2015 at 7:20 AM, Loic Dachary <loic@dachary.org> wrote:
>> Hi,
>>
>> When scheduling suites that are low priority (giant for instance at http://pulpito.ceph.com/loic-2015-03-23_01:09:31-rados-giant---basic-multi/), the --priority 1000 is set because (if I remember correctly) this is the standard priority for nightlies. However, it seems to be different because this morning the giant suite was running although a hammer suite was scheduled before it (for instance http://pulpito.ceph.com/teuthology-2015-03-22_23:18:01-multimds-hammer-testing-basic-multi/) and was still waiting.
>>
>> The giant suite was killed so that it does not interfere with more urgent jobs. Looking at the teuthology jobs there does not seem to be a trace of the priority, probably stored elsewhere.
>>
>> I would be very grateful if you could explain how priorities should be used.
> 
> I've seen a couple of oddities with suites that you've scheduled
> before, and I think something in the system might not be using
> timestamps properly. :) Also, the multimds suite is not an important
> one right now, so feel free to zap it rather than an LTS point release
> run in future.

I would not dare zap/kill a LTS point release run ;-)

> 1000 is definitely the default, and I don't have one handy to look at
> right now but I think if you set a non-default priority it will show
> up in one of the config files.

Right ! It shows at http://qa-proxy.ceph.com/teuthology/loic-2015-03-22_18:50:00-rgw-giant-backports---basic-multi/815885/orig.config.yaml (which was a --priority 101 indeed because it was a single job suite for which I wanted results fast). 

I'll try to re-schedule with --priority 2000 then hoping it will go last. 

Cheers

> -Greg
> 

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: teuthology-suite and priorities

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 1336 bytes --]

Hi Andrew,

I still do not understand what I missed with priorities but I re-scheduled

http://pulpito.ceph.com/loic-2015-03-27_09:03:04-rados-giant---basic-multi
http://pulpito.ceph.com/loic-2015-03-27_09:01:25-rados-giant---basic-multi
http://pulpito.ceph.com/loic-2015-03-27_08:40:06-rgw-giant---basic-multi

with priority 1000. Let see if it runs before the nightlies again or if it behaves ;-)

Cheers

On 23/03/2015 15:20, Loic Dachary wrote:
> Hi,
> 
> When scheduling suites that are low priority (giant for instance at http://pulpito.ceph.com/loic-2015-03-23_01:09:31-rados-giant---basic-multi/), the --priority 1000 is set because (if I remember correctly) this is the standard priority for nightlies. However, it seems to be different because this morning the giant suite was running although a hammer suite was scheduled before it (for instance http://pulpito.ceph.com/teuthology-2015-03-22_23:18:01-multimds-hammer-testing-basic-multi/) and was still waiting.
> 
> The giant suite was killed so that it does not interfere with more urgent jobs. Looking at the teuthology jobs there does not seem to be a trace of the priority, probably stored elsewhere. 
> 
> I would be very grateful if you could explain how priorities should be used.
> 
> Cheers
> 

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]