This is a follow-up topic for Support Case ID 02210537.
We need a way to make Veeam Backup and Replication to be able to ALWAYS copy full VM backups to tapes. Currently, full backups are copied to tapes ONLY IF disk backup job starts and completes the same day, before 00:00. The Support Representative said it’s by design…
At first glance there was nothing special about our backup requirements, it's just a plain Backup-to-Disk-to-Tape scenario. We've created a backup job to backup some VMware VMs, and a Tape job to copy the latest full VM backups to tape. The Tape job is set to run after the source VM-to-Disk job completes.
The thing is we've noticed sometimes VMs were not copied to tapes, with a log message saying there were no new backups to process, which of course was not the case. Furthermore, the job didn’t fail, it completed successfully, but without actually copying anything to tapes. Opening a support case at Veeam resulted in a surprising reply saying it’s by design – and here is the explanation we’ve got: the tape job will copy full VM backup ONLY IF they were taken during the same daily cycle, which starts at 00:00.
There are several issues with such an approach:
1.Backups are usually scheduled to off-hours – and 00:00 is about in the middle of the off-hours period
2.It’s impossible to predict the duration of a backup job, as it can be affected by numerous factors, such as unexpected load on underlying production or backup infrastructure, a CBT reset (say after changing VM disk size) resulting in a full disk re-read, etc.
3.Global organizations might have different daily cycle boundary preference than the local 00:00 time of their backup server
Our current workaround, which is far from perfect, is to split the jobs so those starting before 00:00 will, with high probability (???), finish the same day. The thing is, one would expect a far more deterministic approach to her backup setup… I mean restoring from a backup is your last hope to gain your systems and data back would anything go seriously wrong – and as such it MUST be rock-solid, not depending on uncontrolled external factors. Another issue with our workaround is the backup duration – as described, we had to split the jobs: those that have high probability to finish before 00:00 are scheduled to start at evenings, leaving the longer ones to start after 00:00. As a result, the backup now takes far longer to complete and does not finish on time.
Here are the possible solutions I’ve proposed to the support representative:
1.Add an option to configure the daily cycle boundary, defaulting to 00:00 – this way nothing will change for those not interested
2.Add an option to force the tape copy job to ignore the daily cycle boundary, so it will always copy the latest full backup to tapes – same here, leaving this option off by default means nothing will change for those not interested
The second proposed solution seems to be better than the first one, because it IS deterministic, while the first one IS NOT. The thing is no one can tell for sure a job won’t take more than 24 hours to complete – say due to an unusual load on the underlying infrastructure or a planned change to sizes of several large VM disks – an operation causing CBT reset followed by a full disk re-read – and those are just two of many possible causes for a backup job to take way longer than usual to complete.
Thanks and Best Regards,