gemmellr commented on PR #6036:
URL: 
https://github.com/apache/activemq-artemis/pull/6036#issuecomment-3504940321

   > 
   > Removing this was also the reason why I removed the cleanup job - if we 
let github take care of all the cache, it wouldn't make sense to do the cleanup 
manually ... well, at least if we trust it. I'll take a closer look at what's 
being cached.
   > 
   
   The cleanup is in there for a reason I assure you (I'm the one that put it 
there). Without it, it caches the build output, which is being installed as 
certain tests need it to be installed. Not having the cleanup can also and 
indeed has screwed us over in the past before it was done, more than a couple 
of build issues were masked due to it. Some then only getting hit at release 
time.
   
   I'd personally still want to use the cache action instead of the setup-java 
based stuff regardless.
   
   > > The maven download logging being there is actually desired
   > 
   > I'm just beginning to play with this. I would ask - what's the minimum 
amount of logging that's desired? 100k lines in the log, mostly due to 
downloading, isn't great ... Actually it doesn't even load in the browser for 
me. I think removing the progress is a great way to keep the log smaller, while 
keeping the essential info. The next step, I'm thinking, is to suppress "[INFO] 
Downloading from ..." while keeping "[INFO] Downloaded from ..."
   > 
   
   You'll only get extensive download logs if you have no cache. The main repo 
CI always has a cache. Forks will have one so long as its not your first 
activity in a week....or you do something like removing the existing caching 
stuff and thus forcing it to start afresh hehe. Easy enough to download the 
logs and inspect (though theyve opened in the browser for me, even in the past 
when we drove the significant test output to the console)
   
   Its also desirable to see the 'Downloading from' logs as well as the 
'Downloaded from' ones in either case. I have on multiple occasions found 
repository config issues specifically due to those logs. In a separate case a 
~terabyte of BW may or may not have been wasted over time due to all of the 
downloads logs being disabled hehe.
   
   > > various modules will run tests that cannot run at the same time, so that 
seems unlikely to be too reliable.
   > 
   > Got there! ...
   > 
   > > -T1C
   > 
   > I ran nproc to check how many cores there are available, and the answer is 
4. So theoretically this should help. But. I tried this with builds as well and 
it doesn't look like it's making a world of difference, plus ... it's the tests 
that take longer like you said.
   
   Yep its 4, has been for the last few years:
   https://docs.github.com/en/actions/reference/runners/github-hosted-runners
   
   -T1C does make a noticable difference with up to around 8 threads, but its 
not particularly useful if you arent actually waiting on it, which we arent 
really here. The tradeoff is also exceptionally ugly output (without going to 
things like mvnd), especially on failure (when we most need it), which for this 
job isnt a tradeoff I think is worthwhile.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact


Reply via email to