Last modified 12 years ago Last modified on 08/04/09 18:35:11

dCache Tier I meeting August 4, 2009

[part of a series of meetings]

Present,Gerd,Owen), IN2P3(), Sara(Ron), Triumf(), BNL(Pedro), NDGF(Gerd), PIC(Gerard), GridKa(Silke, Doris), Fermi(), CERN()



(see box on the other side)

Site reports


  • removal of files no longer in pnfs but whose file's state on pool is precious RT4697
  • cleaner has a lot (91) of "failed.pool" files in the pnfsdb/trash/2/current due to the fact we've removed many pools (need instructions how to clean and prevent this so we can improve our retirement procedure)
  • Phenix experiment needs the dcap client for Solaris (follow-up)
  • help on admin console is good... but if the list of commands came in alphabetical order it might be better (don't know if, right now, there is any logic on the order).


Gerd reported that they have now upgraded NDGF back to v1.9.4-2. They had previously downgraded from this version because it was believed that v1.9.4-2 suffered from the database leak problem. In fact, the leak problem was not present in 1.9.4-2, but that a database clean up was necessary.

After the upgrade, things seem fine.


Doris reported that they have downgraded to dCache v1.9.1-5 & v1.9.1-6. The systems are now stable.


Things are fine at PIC. They have experienced no problems due to the database leak. They are planning to upgrade tomorrow to new version: v1.9.2-10 for SRM. The pools will remain at 1.9.2-5 and the remaining head nodes at 1.9.2-7.


Ron reported that they upgrade to dCache v1.9.3-3 last week. They have seen no issues so far.

Postgres 8.4

Doris asked: has anyone tried it yet?

Owen reported that dCacheConfigure hasn't been made to work with this yet. This is his 2nd priority.

Pedro reported that BNL are planning to move to PostGreSQL v8.4 in September. The motivation comes from the new features that PostGreSQL v8.4 brings; specifically:

  • Auto-Tuning Free Space Map,
  • Improvements in autovacuuming

Someone asked Pedro if he plans to take advantage of new ability to create a read-only copy of the database?

Pedro replied that BNL currently uses the PITR support so could replay the WAR files to achieve the same effect.

Paul mentioned that DESY are using Slony to achieve hot-spare machines. This could also be used for off-loading read-only queries.

Gerd said that, with v8.4, they is better support for ?? and, with v8.5, they plan for add more support for ???

Doris is planning on using v8.4 for new dCache CMS instance. They have been using an older version so far, but since the instance currently has no data wiping the databases and using PostGreSQL v8.4 shouldn't be a problem.

Http domain

Doris asked about whether doing a reload on the web-browser works. In particular, what is the update frequency for the information in the web pages?

Gerd reported that (most) pages within the web interface are regenerated every 2 minutes.

Doris' question was targeted at the tape restore monitor. FZK would like to monitor whether there are files that cannot be restored from tape. They could achieve this via the admin interface or from scanning the web page.

Gerd said that he thought the tape-restore is a special case: the page is generated on-demand.


Same time, next week.