After running so reliably for such a long time, I’d received another e-mail from CrashPlan today to say my backup set was below 100%.  Time for some investigation….

The last time, the drive was full, so I needed to expand the VirtualBox Virtual Drive which I blogged about here.  No space issues this time, but when I checked the VM Crashplan wasn’t running.  I started it up, and it started doing its thing, and then went back to check 30 minutes later and it had shut down again.

There were a few CentOS updates that needed applying, so I thought I’d run those and then restart and investigate further.  After the updates, I launched CrashPlan again and kept the window open on the screen, and sure enough, after a few minutes, the application crashed.  I’ve been using CrashPlan for some time and was aware that it needs more memory to run the larger the backup set becomes.  Mine is now just over 4.2TB with over 6 million files, so I’d allocated 4GB RAM to the VM.  I thought maybe it was bouncing on this limit, so shut the VM down and opened up the VirtualBox WUI to check the settings.  Not quite sure how or why, but the memory allocated to the VM was set at 2048, which is not enough for Crashplan to run happily.   I bumped this back up to 4096 and then restarted.

I was expecting this to work fine, but it didn’t!  Same issue, with the Crashplan opening and then closing down within a few minutes.  I shut down the VM again and bumped the memory up to 5120 (5GB) and tried again, but the same issue.  Time to check out the CrashPlan support website.  I found the article I was looking for which had the details about the recommended settings based on the size of the backup, but also had some further details about how to troubleshoot.  The recommended solution is to set the java memory allocation, which I did by:

# double-click the Crashplan logo, which bring up the Help window
# enter 'java mx 4086, restart' into the command line at the top

Crashplan started back up, and for the last 30 minutes at least hasn’t shut back down.  Not sure why the VM memory changed, but I suspect this reset the java value I’d previously set in Crashplan, so hopefully now everything will be fine again for some time.  Fingers crossed….