Version 1 (modified by paul, 3 years ago) (diff)


dCache Tier I meeting MONTH DATE, 2013

[part of a series of meetings]

Present, IN2P3(), Sara(), Triumf(), BNL(), NDGF(Jens), PIC(), KIT(Samuel), Fermi(), CERN()


(see box on the other side)

Site reports


Things are pretty happy.

sweeper purge DoS attack

We had an issue yesterday.

Benchmarking -- checking pool start-up time.

Many small pools: tried 1 pool, 5 pools, 20 pools

Yes, this greatly improves start-up times.

When the test was over -- "sweeper purge" to clear cache.

Unfortunately, "sweeper purge" lead to severe impact on dCache .

Central services were affected, complaining about PNFS messaging timeouts (e.g., PinManager? complaining).

After 10 minutes, dCache recovered.

Last configurable was with 20 pools.

There were 850,000 files in total

Also found some stack-traces. Will report these.


Everything is running fine.

ZK issue

Currently running stand-alone zookeeper cluster with three nodes. We want to move one ZK server to another rack; in effect, decommission one zookeeper server and commission a new ZK server.

Currently, this requires restarting all dCache domains, since the ZK connection string is changing. Could we use a DNS aliases (one for each ZK server) to avoid having to restart dCache domains?


Accessing the file twice with the same UUID.

Host running test every 15 minutes.

Noticed changing the xrootd debug level from 2 to 3, now seeing the UUID

Support tickets for discussion

[Items are added here automagically]


Same time, next week.