Jump to content


  • Content count

  • Joined

  • Last visited

Everything posted by dfoster

  1. Knowing that Retrospect was RAM hungry, the servers have 4GB each and should not be paging too much (our largest catalogue is about 1.5GB) Daniel.
  2. I noticed dueing an manual restroe that Retrospect never used more than 50% of the available processor capacity during the matching process. It is a dual processor box and so my first though was that it was just using the one processor. I was surprised to discover that this was not the case. Both processors were being used and the utilisation graphs were almost mirror images - as on prcessors utilisation increase the other dropped and the combined result was always 50% utilisation or less. As this is a dedicated backup server it would be preferable to use the entire available capacity - anyone have any clues? Cheers, Daniel Foster
  3. Doesn't appear to be controller related - swapped controller cards, PCI slots Drive enclosure and still get said errors - very frustrating. Also tried firmware updates, resricting the size of requests to 128K (an MS suggestion) with no result. Google-ing the error shows that it is actually a pretty common problem and that there is no single solution. In my case it looks related to bus utilisation in that the error rate increases when more than one drive is being written to at once. Time to try a different server - maybe the PCI bus on the Sun is the source.... Daniel.
  4. For some reason I keep getiing write failures to our external firewire drives when there is no user logged on the the box. If a user is logged on (either via the local console or even a disconnected terminal session) the backup work perfectly, but if the box is left without a user logged in, I get the write failures again which corrupt the backup set. I have checked the drive, updated the bridge chip firmware and swapped controller cards to no avail. Seems to be something odd the retrospect is doing. While I can work around by ensureing that there is always someone logged on, this is not ideal as it means that the backups are corrupted if the box restarts itself (crash, power outage etc) and no-one gets to it before the scripts run. Daniel. ------ Daniel Foster Systems Administrator Faculty of Arts, University of Western Australia.
  5. Needing a users logged it seems now to have been a red herring - we are now getting write errors both whether logged in or not (I guess we were just getting luck in earlier testing...). As a result it now looks like a Windows issue rather than a Retrospect one. This has been confirmed through some large file copies between drives which can be used to replicate the error. Now to try and get a response from MS - I may get lucky.... Mostly we are getting sbp2port errors stating that " The device, \Device\Sbp2\Oxford Semiconductor Ltd. , did not respond within the timeour period." We also get the occasional FTdisk error "The system failed to flush data to the transaction log. Corruption may occur." The result in retrospect is that the execution unit hangs and in rare cases, the FTdisk error results in the entire disk being hosed. We have tried having the drives set for both performance and quick remove - neither setting affects the rate of errors. The drives are formatted with NTFS (they are 250GB drives) so Windows complained about having the quick removal setting on.