09 May 2017, 14:00 - 15:26
Present: Oxana (from airport), Maiken, Mattias, David, Aleksandr, Anders
Apologies: Balazs
== News
- Mattias did some arc.conf work, checked in on svn, however not yet documented in the tracking file (https://docs.google.com/spreadsheets/d/14Abr9J5YxukDLaqo8uNmyJDJ138NmtpdmLiYHrtzH5w/edit) (nor arc.conf.reference in the svn branch). There was some confusion (Maiken) how the changes really should be documented. What is the purpose of the tracking file versus the notes and change-tags (e.g. #91 CHANGE) made in arc.conf.reference? Need reminder, or some text explaining the wanted procedure on the arc.conf wiki. (http://wiki.nordugrid.org/wiki/Arc.conf_review)
- On the same note: noticed the link in the wiki to svn arcconf_restructure branch is wrong (typo)
- David: He has a test-CE that keeps wanting to update to 5.3.0.
- Reopened bug 3641 - SSL errors with CERN certificates. Found another case of it. Aleksandr is on it.
- Tried to install A-rex without ldap as he is setting up an EMI-ES CE and does not need ldap. Does not work. Discussed below.
- Oxana: Cern Comitee meting for Tier 1. ARC instability mentioned by HEPIX - meeting of all sys-admins. Several people reported that ARC is "bad". Question is: is there a general feeling that ARC is bad? David has not heard anything like that from the ARC pilot-sites. Oxana should try to find out a bit more around this, where does the rumour come from, what is the main reason? Is it just that ARC is used by more and not compared to CREAM anymore? Suggestion: Send developers to HEPIX. Or arrange some user-forum for questions/discussion.
- Aleksandr: has been busy with code fixes and enhancements.
- Anders: working on the nightlies build system. Managed to get it up and running and produced nightlies 2017-05-08. Still needs more work since 2017-05-09 are not produced. But going in the right direction.
- Have been working on setting up python bindings on the compute client. Non-trivial to submit a job purely from python. Would like to start from python rather than from xrls.
- Maiken: Abel ce01 has installed 5.3.0+memory fixes. We have determined that the critical fix was commit 36357 by installing 5.3.0+memory-fixes except 36357 which resulted in similar behaviour as 5.3.0 although took a bit longer for a-rex to really come to a halt (probably since there were after all other memory leak fixes). However, the AnonPages just kept increasing and increasing like they did with original 5.3.0. Triolith has installed 5.3.0+memory fixes. However, Triolith with 5.3.0+memory fixes encounters restart from watchdog pretty much every half hour or so, much like what we saw in original 5.3.0. Is under investigation by Jens and Aleksandr.
== Release
- As 5.3.0+memory fixes installed on Abel ce01 showed good behaviour, we push 5.3.1 release with the memory fixes-> 5.3.1. Will also include nagios probe commits which will solve ggus tickets 126724 and 127026, and include small fix for bug 3654 - Source files for arctst -J 1 unavailable from ftp.nordugrid.org
- After this we will push for a 5.3.2 release not too long after, with further enhancements.
- Also for next major release: work is needed to detangle ldap and arex further. Possibly also python-bindings. (Bug 3653)
- Configuration restructuring continues in background.
== Bugs
- 3635: Not trivial to detangle A-rex and ldap possibly due to Configparser.pm. Python-bindings should also be investigated. Needs some dedicated work. <Anders, David, Mattias, please fill in some details>. Should be resolved before ARC 6.0.
- 3654: Anders fixed in trunk.
- 3641: Aleksandr is looking into it.
== AOB
- None.
There are minutes attached to this event.
Show them.
