Part of storage system Pica is still very slow


We are sorry to say that part of storage system Pica is still very slow. For some users this makes Milou for the moment almost of no use

Tomorrow Thursday UPPMAX will look more into the problem.

Most probably this is due to some user that is loading down Pica in a bad way, but probably unintentional.

If you think that you might be this user, please think if you are using Milou (or perhaps Rackham, if you see the same problem there) in a new or unusual way. If so, please e-mail us at our support address, and we will try to adapt your jobs, so they work better.

This might happen if  you for example:

  • Run several jobs in the same directory
  • Do a lot of rsyncing or gzipping on Pica
  • Have several jobs that do lookup in file based databases

I will update this newspage, when I have more information about the problem.

This is vacation time at UPPMAX, so Thursday we are only me, on Friday we are none, and next week we are only two. So I hope that something can be helped here soon, but it more difficult than usual to fix things.

Update Thursday at 1030 hours

Now the situation looks much better. The read and write traffic to Pica runs much more smoothly. I guess that this is due to some jobs, that were unfriendly to Pica, now has finished.

We might of course get the problem back. Please be careful. If you notice any problem, and you have many jobs running, it might in that case be your jobs that create the problem, if you are unlucky, and you might try to have fewer jobs running at the same time.

You are welcome to report your problems to us. Most valuable for us is to know when problem starts, when problem ends, and where you have your files

System News