Welcome!
- the current release is 1.8.7. We finished pushing the packages to EPEL-test. Only dpm-xrootd is giving bureaucratic issues.
yaim, metapcakages and oracle things should go to:
- emi testing as a testing platform
- wlcg as final destination
- moving to component-based releases. The scenario will become more tricky to understand, as we will not have "DPM releases" covering everything as we had so far. This is an evolution of the versioning practices of the whole lcgdm project
- When to use only Puppet in our trunk and RC testbeds ? What's the future of yaim?
- date to stop yaim support
- propose Aug 2014 as the stop for yaim support
- period during which to upgrade
- where should we publish the puppet manifests?
martin: yaim is not supported anymore from emi, our components are still configured with yaim. The proposal is a grace period, we start supporting puppet and we end supporting yaim after some time. We propose supporting yaim until aug2014, to give enough time to learn puppet and to give a time frame that makes us sure that everyone has enough time before the lhc restart. We support both small setups (one-node only) and bigger ones with the puppet master.
wahid: expect a deluge of questions. uk sites were already starting to ask things.
oliver: is there anything obvious that we have not done? wahid: not really. it also depends from how confident you are
martin: proposal is "go for it with the helmet", don't run it immediately on a prod site, puppet has to be understood a bit, not used blindly. We are happy to help anyone.
a.Doria: will it involve all the services of a site at once? Oliver: not necessarily, it's a site choice. In any case the minimal investment is for the sysadmin to learn it in the standalone mode.
sam+oliver: we should provide adequate docs. the push to puppet should not be interpreted as forcing everybody
- manifest distribution: two possibilities: rpm and github
- rpm would contain the dpm manifests, and also other ones, e.g. mysql. adrien remarked that there are no other similar examples. people use github. what do you prefer?
wahid: there are both cases. some sites are used to github, others not.
valtri dvorak: the rpm is uncomfy because it prevents relocation. proposes to follow the rules of the puppet labs sites
oliver: is it a disaster for anybody if we do not support the rpm? silence... seems not a disaster
- Going towards IPv6. what is the status?
apache ok, dpm/dpns ok
sam: member of the hepix ipv6 group. dpm at glasgow uses ipv6 and it works. yaim gets confused with ipv6 and needed hacks. mysql has issues because the one in the distros is not new enough, need 5.5.
rfcp seems ok, gridftp seems ok with some hacks. xrootd no. webdav seems ok. generally it's a pretty good situation.
adrien: what about srm? sam: seems ok!
oliver: could you submit from now on tkts for the issues that you found. ipv6 support from now on becomes official thing. we will fix every bug you open about ipv6 because it's valid stuff
glasgow is one of the sites who did ipv6 testing for hepix
dpm seems fine. it's what is attached to dpm that may give issues
- eventual followups about NFS and VFS
nfs: metadata ops are fine, now getting to the new layout
oliver: is it part to moving to dmlite? yes. timescale?
ST: we plan to release by september, and commit to the original CERN repo.
andrea: rw or r/o... r/o
adrien: do you plan epel? ST: it's maybe late. oliver: when the time is right we can help.
fab: is it ok if we forward the tkts to you? ST: yes
NFS czech.... oliver: please send us info or we can setup another meeting
VFS czech: postponed.... mickey mouse problem.
- Wahid is starting to run regular performance tests (ROOT? Hammercloud?) towards the release candidate testbed
wahid: we can use the test yes, we can test wan access as well.
the tests can be very heavyweight, depending on various things
adisalvo: it would be good to have the same testing level also for the testbeds in italy. Wahid: we can get in touch and get integrated
- xrootd perf and http. First attempts works, it's largely suboptimal
- status test deployment of the ita sites?
dpm test endpoints: a.doria will send email in the afternoon
federations: want to use xrootd+https feds for POD testing
a.doria: we take as much as possible from epel test. dpm-xrootd is missing
- Organising the support rota
- What does the responsible person do: answer and set in progress snow tickets
- Monitor the builds and write to dpm-devel if anything is broken
- Monitor the tests (when we have them) and write to dpm-devel if anything is broken
oliver: uk and france will join such a rota
fab: circulate mail to gather names 360 degs
- Choosing a regular slot for the next meetings.
fab will set up 1 year of mtgs, last friday of each month
- AoB
switch rc testbed to puppet only ... since now!
Cheers
Fabrizio