Rare project developers

[Alexander Gall <>]

On Tue, 7 Feb 2023 09:03:02 +0100, mc36 <> said:

> hi,
> i've commited today morning to the bitbucket
> https://bitbucket.software.geant.org/projects/RARE/repos/rare/commits/cc7cfe6a8f3b78e6f64b884c2001daa7b82fff1e
> then i updated the optimize_profiles tag but seemingly nothing happened...
> could you please give me hints about what i missed here?

The webhook terminates with a 502 error if you trigger the process
with the tag. I don't understand why. I just did the same and it
worked (the script still failed because the VM took too long to start,
maybe I should increase that timeout).

> ps: is it possible to have a /rerun or something command to the slack bot 
> for such cases?

Just remove and re-add the tag. I have enabled some debugging to see
why it fails when you do this.

-- 
Alex

> thanks,
> cs


> On 1/16/23 15:34, mc36 wrote:
>> you're unbelievable, thank you so much! :)
>> 
>> On 1/16/23 13:55, Alexander Gall wrote:
>>> Hi csaba
>>> 
>>> I have implemented the mechanism for automated profile optimization I
>>> described earlier. You can use it simply by pointing the Git tag
>>> "optimize_profiles" to the commit of the RARE repo for which you want
>>> to perform the optimization. Progress is reported live on the Slack
>>> channel https://wp6t1-wp-rare.slack.com/archives/C04JM1LND63. I took
>>> the liberty to invite you (and Fr  d  ric) to the channel. I know you're
>>> not a Slack person, but you only have to check that channel after you
>>> initiated an optimizer run :)
>>> 
>>> The optimization finishes either with
>>> 
>>> Status: Success
>>> 
>>> or
>>> 
>>> Status: Failure
>>> 
>>> In both cases, the output of the optimization process is uploaded to
>>> the channel. You can request the output for a running optimizer with
>>> the "/progress" command in the channel at any time.
>>> 
>>> The "Success" status means that the optimizer was able to find working
>>> values for all of the profiles, where as "Failure" either means that
>>> at least one profile could not be optimized or something else went
>>> wrong. The log file should say what exactly the problem was in that
>>> case.
>>> 
>>> For "Success" there are two possible outcomes. If none of the
>>> optimized values actually changed (compared to the tagged commit),
>>> nothing else happens and the log ends with something like
>>> 
>>> No changes detected
>>> 
>>> This means that the RARE commit tagged with "optimize_profiles" can be
>>> safely added to the RARE-NIX master branch (in rare/repo.nix).
>>> 
>>> If some of the optimized values actually changed, the optimizer adds a
>>> new branch named "optimizer_<sde-version>_<random>" in the RARE repo
>>> and creates a PR to merge it to the master branch. One of us can then
>>> review the generated code and merge it. It is recommended to use the
>>> option to automatically delete the branch after the merge (the
>>> <random> part of the branch name is there to avoid a conflict in a
>>> future run of the optimizer in case one of the old branches is still
>>> around).
>>> 
>>> Then, finally, the commit that results from that merge can be added to
>>> the RARE-NIX repo (rare/repo.nix).
>>> 
>>> I have added the tag this morning to
>>> https://bitbucket.software.geant.org/projects/RARE/repos/rare/commits/554626230089032b5065d6b6749115dca3642597
>>> 
>>> Here is what happened when I did that:
>>> 
>>> - The webhook "RARE P4 Profile Optimizer" associated with the RARE
>>> repo was triggered
>>> 
>>> - The webhook sent a POST request to
>>> http://hydra.nix.net.switch.ch/cgi-bin/profile-optimizer.cgi
>>> 
>>> - The CGI script checked that the event was the addition of the tag
>>> "optimize_profiles" (any other event is ignored)
>>> 
>>> - The script then activated a VM in the SWITCH cloud and executed a
>>> helper script
>>> 
>>> - The helper script cloned the RARE repo and checked out the commit to
>>> which the "optimize_profiles" tag points and executed the
>>> profiles/optimize.sh script
>>> 
>>> - The optimize.sh script checked out the latest commit of the RARE-NIX
>>> repo and ran the optimizer in the context of that commit (i.e. it
>>> used the SDE version and p4c compiler options as required by that
>>> version of RARE-NIX)
>>> 
>>> - The optimizer run was successful, i.e. it found a working
>>> configuration for all profiles
>>> 
>>> - The optimized profiles did *not* differ from those already present
>>> in the RARE repo tagged with "optimize_profiles"
>>> 
>>> I.e. this was an instance of a successful run that did not produce a
>>> new set of optimized values and hence no branch and PR.
>>> 
>>> Please don't move the "optimize_profiles" tag while the optimizer is
>>> running. The system can't deal well with multiple concurrent optimizer
>>> runs :( It can probably also fail in many other ways, it's all a bit
>>> hackish...
>>>