|Main||EGI.eu operations services||Support||Documentation||Tools||Activities||Performance||Technology||Catch-all Services||Resource Allocation||Security|
Minutes – Grid Operations Meeting – 28 February 2011:
Alexander Kryukov, Anders Wäänänen, David Meredith, Drazen Zujovic, Emrah Akkoyun, Feyza Eryol, Foued Jrad, Mario David, Michaela Lechner, Peter Slizik, Peter Solagna, Stuart Purdie , Tomas Kouba, Ulf Tigerstedt , Vladimir Slavnic, Wen Mei
1 - Information (Mario)
- Staged rollout wiki and procedures updated:
- New Early Adopters portal (updated):
- Work in progress to implement all needed features in the staged-rollout RT queue, for the staged rollout tests
- All components of glite 3.2 are presently covered by EAs, except the tarballs WN and UI, thus need teams to cover them
2 - Staged Rollout (Mario)
2.1 - gLite 3.1
- LFC and DPM 1.8.0-1 : all patches will be rejected since they have voms-api-c(pp) with mem leak, and new ones will be produced
- FTS/FTM/FTA: are in certification.
Mario: contacted the developers, they have put this into low priority, due to work on the EMI release. Ask sites/operations to send feedback and justification to put more pressure on the developers, if sites want this new version.
Peter: a new release implies only the new repackaging of DPM with the new dependency on the voms-api that has the memory leak fixed. Having around 100 sites with DPM in glite 3.1 is already a good justification. Asks how big is the memory leak affecting the service.
Mario: this has been seen in a DPM at an atlas tier2, and the site administrator has implemented a cron to restart some of the deamons. Sites/Operations have to send feedback if they want the new DPM, sites may be waiting for the EMI release, and may be happy with the stability of the current production service.
2.2 - gLite 3.2
- Ready for Certification, in Certification - Upcoming in the next few days to staged rollout:
- L&B 2.1.10
- Torque (utils, server, client)
- UI 3.2.9
- VOBOX 3.2.11 - Went into staged rollout already
- FTA/FTM/FTS 2.2.5
- In staged rollout - both set into ready for production, will be released in the next couple of days
- WN 3.2.10 : No major problems found in staged rollout
- glite-CLUSTER 3.2.2 : No major problems found in staged rollout
- Urgent fix in CREAM - fast staged rollout test, already released into production:
- Problem reported by LHCb with CREAM 1.6.4 - there was an urgent fix that went through staged rollout in a very short time (less then 1 day). New version CREAM 1.6.5
2.3 - Operational Tools
- Staged rollout of SAM update 9 https://rt.egi.eu/rt/Ticket/Display.html?id=1281
- Problems found in the test, in contact with developers.
3 - Operational Issues (all)
3.1 - IGTF CA 1.38 and VOMS-ADMIN (Mario)
Problem found in production VOMS-ADMIN server after update to the latest IGTF-CA 1.38 with the new egi policy format. A Post-Mortem analysis is in progress and will be documented soon.
3.2 - Where to report problems/issues for SW in staged rollout (Mario)
All SW components under staged rollout, if problems are found they should be reported always through GGUS, either if there are SLAs with the Technology providers, or if they are developed and supported in a best effort through the community. Relate to the problems found in the latest SAM/Nagios update that the EA has found problems and opened a GGUS ticket, but in the end contected directly the developers.
3.3 - TopBDII has http://goc.gridops.org/gocdbpi/... hardwired, but this will be decommissioned (Mario)
- Followed in GGUS ticket: https://gus.fzk.de/dmsu/dmsu_ticket.php?ticket=67544
- The new endpoint is: https://goc.egi.eu/gocdbpi/...
- Discussions and progress with the developers on one side. Ops tools and Operations on the the other side
3.4 - EGI BROADCAST spam must be culled (Ulf Tigerstedt, NGI_NDGF and NGI_FI)
As everything gets spammed with EGI BROADCASTs nowadays, the people getting them in 10-folds are beginning to treat them as spam. Could the system be improved to not spam every mailing list available for every single thing? As a ROD I'm currently getting 4 copies, some others are getting 14...
From Cyril Lorphelin
Hi Mario , Ulf
This problem is quite difficult to solve. We propose a tool which permits to send emails widely. If people don't choose properly the target everybody will be informed .
We have added the option to create models // templates with specific subjects and specific targets , this is probably a solution to reduce spams . It means that people should properly select the right targets when they are creating a model .
I have no obvious solution.
Ulf: some sites (specially arc) are getting to many EGI broadcasts and treating them as spam, they are not interested in glite release announcements for example.
Mario: need suggestions and proposals to try to minimize this, and target better the several broadcasts.
Action on Mario: to open a RT ticket to gather suggestion/feedback on this issue.
4 - AOB
Michaela: reminder that today is the last day for early bird registration into the EGI User Forum.
Next Grid Operations Meeting:
14 March 2011, 14h00 Amsterdam time