Denis wrote:
> As I understand it, slave delegation
means the master is tossing Java objects around to the slave for execution,
so the master doesn't just sit idle while the slave does its work.
Do we have evidence that this is really
true? It wouldn't make sense at all. Aren't our tests one monolithic application
that is started by a shell script and that executes purely on the slave
(starting secondary JVMs as instructed by the test.xml ant scripts)?
Markus
From:
David M Williams <david_williams@xxxxxxxxxx>
To:
Cross project issues
<cross-project-issues-dev@xxxxxxxxxxx>
Date:
2013-10-24 18:00
Subject:
Re: [cross-project-issues-dev]
Shared Hudson instance reacts extremly
slow
Sent by:
cross-project-issues-dev-bounces@xxxxxxxxxxx
I wonder if the Hudson project has any advise/opinions on if an upgrade
to Hudson would help reduce "interference"? We are on a pretty
old version. Bug
412933 - Should we Upgrade to Hudson 3.1.0?
Put another way, if we can't run a Mac Machine and a Windows Machine as
two independent slaves ... Hudson is not doing us much good :)
And to put yet another way ... if we had to run them serially ... our 6
to 8 hour test suites would seldom ever finish! :(
Thanks,
From: Denis
Roy <denis.roy@xxxxxxxxxxx>
To: cross-project-issues-dev@xxxxxxxxxxx,
Date: 10/24/2013
08:46 AM
Subject: Re:
[cross-project-issues-dev] Shared Hudson instance reacts extremly slow
Sent by: cross-project-issues-dev-bounces@xxxxxxxxxxx
On 10/24/2013 02:40 AM, David M Williams wrote:
> I'm not sure it's wise to have the ep4-inut-mac64 and ep4-unit-win32
jobs both be running at the same time, though....
I don't understand why this would be the case. The are separate physical
machines, right?
They are separate physical machines, but they are slaves. As I understand
it, slave delegation means the master is tossing Java objects around to
the slave for execution, so the master doesn't just sit idle while the
slave does its work.
But, back to the point of original topic, it is not so much that a "sudden
restart caused [us] some grief" but the fact that there is no clue
or announcement that it was done, which is the point of Bug
417457.
So here we are, providing some clues.
Matt is restarting the master now, since it has been tied up all night.
I think it would be a great idea to avoid running both of those jobs
at the same time to see if it takes care of the non-response issue.
I'm not sure it's wise to have the ep4-inut-mac64 and ep4-unit-win32 jobs
both be running at the same time, though....
I see 36m left for win32, 2h05 left for Mac.
Denis
On 10/23/2013 04:57 PM, Denis Roy wrote:
I think the Windows/Mac tests seriously affect the master performance.
Last time we've seen this David was running tests and our sudden
restart caused him some grief.
On 10/23/2013 04:24 PM, Denis Roy wrote:
Yep, but it's running eight builds right now, so we don't want to restart
it.
Hang in there...
Denis
On 10/23/2013 04:17 PM, Tom Schindl wrote:
... anyoneelse seeing hudson being extremly slow?