Scratch overburdened

Dear HPC team,

I currently have a workflow that performs i/o on scratch on Baobab, but it looks like the scratch is quite slow at the moment. Even ls and cd take a while.
Can something be done regarding this?



Hi @Debajyoti.Sengupta

Could you please give us more details, on wich cluster do you have this issue ? which directories ?

Hi Adrien,

Apologies for the lack of information.

This is on baobab. And the workflow needs access to /srv/beegfs/scratch/groups/rodem/skycurtains/fullskyscan/

I am finding the same on baobab. Anything on scratch, my own space or on /srv/beegfs/scratch/groups/rodem/ takes forever. As Deb said, just running ls takes a while.

I am trying to clear up a lot of space by deleting old runs, but the issue persisted before. This morning many of my jobs would fail after timing out trying to save simple log files.

1 Like

I am having the same issue, everything on scratch is slow

Hi @Matthew.Leigh

I am looking for it but it’s quite difficult to find the root cause.

Dear users,

I remind you that we have now a new cluster: Bamboo.
This cluster have a much faster storage! Home is fully on SSD.
Maybe a good time to give a try?!