Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 442453

Summary:	Need /shared/eclipse defined on hudson-perf1master
Product:	Community	Reporter:	David Williams <david_williams>
Component:	CI-Jenkins	Assignee:	CI Admin Inbox <ci.admin-inbox>
Status:	RESOLVED FIXED	QA Contact:
Severity:	major
Priority:	P3	CC:	denis.roy, webmaster
Version:	unspecified
Target Milestone:	---
Hardware:	PC
OS:	Linux
Whiteboard:
Bug Depends on:
Bug Blocks:	374441

Description David Williams

2014-08-25 01:18:37 EDT

I mentioned this in bug 374441 comment 44 but may have gotten lost there. 

There "dedicated performance machine" apparently has /shared/common defined, but, not /shared/eclipse. 

It's getting close to the point this will be "blocking", so was hoping to get it resolved before then.

And, by "resolved", I mean either have it defined, or explained why it can't be and what else is recommend? 

The main purpose we need write access from hudson test machine to "/shared/eclipse" is primarily because we write a small file to a well known directory there, and a cron job looks for that file (every 10 minute or so) then knows to begin the process of fetching the data from Hudson, doing some analysis, and, eventually, putting results on 'downloads'. 

If for what ever reason you can't define it that way, then it would be a major effort to figure out how else to do it (plus, then we'd basically have one method for unit tests, and another for performance tests ... which just complicates things if we can not use "common code" for  parts of it.

Comment 1 Eclipse Webmaster

2014-08-25 15:25:50 EDT

I've taken a look at hudson-perf1-tests(after poking it) and it seems to be mounting the 'shared' area in the exact same way as the master, and I can clearly see /shared/eclipse.

Is there something I'm missing?

-M.

Comment 2 David Williams

2014-08-25 16:50:29 EDT

(In reply to Eclipse Webmaster from comment #1)
> I've taken a look at hudson-perf1-tests(after poking it) and it seems to be
> mounting the 'shared' area in the exact same way as the master, and I can
> clearly see /shared/eclipse.
> 
> Is there something I'm missing?
> 
> -M.

Well, I was going to re-run my test job that showed the error originally, but it now says it's down. Getting "503". 

That's at https://hudson.eclipse.org/perftests/

Comment 3 David Williams

2014-08-25 17:33:28 EDT

(In reply to Eclipse Webmaster from comment #1)
> 
> Is there something I'm missing?

I see. We are talking about two different machines. I see 'huson-perf1-tests' as a slave on the "normal" shared Hudson instance ... it has no jobs associated with it currently, but apparently we ran a few there, 1 year and 11 months ago. 

And, honestly, don't recall what that is, or why we were given "another" machine ... hudson-perf1master ... except that it was "more dedicated" to just our performance test ... or something? Just a "master" machine, skip all the slave noise? 

So, poke at that one a while and see how 'shared' is defined :) 

I'm assuming the two are unrelated?

Comment 4 Eclipse Webmaster

2014-08-26 10:15:21 EDT

Ok my bad.  I shutdown the perftests machine so I could start the actual slave(since I thought that's what you were using).  I've shutdown the slave, restarted your instance and updated /shared so that it points to the mount point which means you can now access /shared/common and /shared/eclipse .

-M.

Comment 5 David Williams

2014-08-26 10:41:14 EDT

(In reply to Eclipse Webmaster from comment #4)
> Ok my bad.  I shutdown the perftests machine so I could start the actual
> slave(since I thought that's what you were using).  I've shutdown the slave,
> restarted your instance and updated /shared so that it points to the mount
> point which means you can now access /shared/common and /shared/eclipse .
> 
> -M.

ok, thanks. It is still showing "service unavailable" (503) to me ... at 10:40. 

I think that's enough time for it to come up. So maybe went right back down?

Comment 6 Denis Roy

2014-08-26 11:06:48 EDT

> And, honestly, don't recall what that is, or why we were given "another"
> machine ... hudson-perf1master ... except that it was "more dedicated" to
> just our performance test ... or something? Just a "master" machine, skip
> all the slave noise? 

I wasn't sure if performance tests would be reliable when performed on a slave machine.  I know Hudson does delegation at the object level, so perhaps that delegation would cause jitter in the results.

The idea behind a dedicated slave was to avoid those fluctuations.

Comment 7 David Williams

2014-08-26 12:33:23 EDT

(In reply to Denis Roy from comment #6)
> > And, honestly, don't recall what that is, or why we were given "another"
> > machine ... hudson-perf1master ... except that it was "more dedicated" to
> > just our performance test ... or something? Just a "master" machine, skip
> > all the slave noise? 
> 
> I wasn't sure if performance tests would be reliable when performed on a
> slave machine.  I know Hudson does delegation at the object level, so
> perhaps that delegation would cause jitter in the results.
> 
> The idea behind a dedicated slave was to avoid those fluctuations.

Yes, I think this is a good idea. (As I recall, we were seeing a lot of communication between master and slave, just to "check status" or something. 

Gut, in case it was missed, it is still down, if the URL is the same: 
 
https://hudson.eclipse.org/perftests/

- - - - - - 
Service unavailable!

The server is temporarily unable to service your request due to maintenance downtime or capacity problems. Please try again later.

If you think this is a server error, please contact the webmaster.
Error 503
hudson.eclipse.org
Tue Aug 26 12:28:56 2014
Apache

Comment 8 Eclipse Webmaster

2014-08-26 15:39:55 EDT

I've started the Hudson instance on perf-master.

-M.

Comment 9 David Williams

2014-08-26 15:42:17 EDT

Thanks, I confirmed I can see it, and my little "peek test" can see /shared/eclipse ... so, I'm set.