CVMFS link very slow

HPC Technical
1

Hi,

recently the cvmfs link on both Yggdrasil and Baobab is really slow.

Is it possible to restart the corresponding server? or is it linked to a slow connection to CERN?

Thanks and cheers,
Manuel

1 Like
2

Hi, can you please give more insight ?

How are we supposed to check the perfs? Do you have the issue on a specific file?
Are you accessing the data from the login node or from a compute node?

3

Hi Yann,

Unfortunately we don’t have too much extra information.

We are having the issue when accessing any resource from the cvmfs mount, not being familiar with how the SQUID is set up (and I’m not sure even where this is run!) we’re not sure if this is a bandwidth or latency issue on our CERN link, or something with how it is configured.
However in trying to source a bash script on the cvmfs server the connection hangs or runs very very slow. This is on both login and computing nodes.
The command I need to run first before using any cvmfs resource is

export ATLAS_LOCAL_ROOT_BASE=/cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase
PKG_SETUP=${ATLAS_LOCAL_ROOT_BASE}/packageSetups
source ${ATLAS_LOCAL_ROOT_BASE}/user/atlasLocalSetup.sh

I know this was something set up with gianfranco.sciacca at Bern.

Cheers,
Johnny

4

Hi Yann,

could it be related to the latest comment on this post from David, that the DPNC Beegfs server is not mounted.
Is the link to the DPNC cluster up and running? As I recall it is over this link that we have our direct access to CERN.

Cheers,
Johnny