When using Integrated Storage and Windows VSs with large vdisks, sometimes the Cloudboot Backup Server might crash during backups.
This issue is related to the fact that ntfsclone does not handle read IO errors properly during data being copied off a vdisk snapshot. The read error is generated when the vdisk snapshot reservation space is exhausted during backup, due to a large amount of IO writes during the backup process or due to a long backup process on slow storage devices.
If you have Windows VS and you frequently experience Backup Server crashes or backup operation failures, you might consider implementing these solutions:
1) Scheduling the Windows VS backups during the idle times, when there are not many IO writes to the live vdisk. Disk IOPS can be viewed in the vdisk menu of the VS.
2) Increasing the snapshot reservation space for the vdisk(s) on the Backup Server CLI:
onappstore snapreserve uuid=<UUID> reserve=<SNAP_RESERVE_IN_PERCENTAGE>
The default percentage is 10%, and it can be increased to a desired value for the specific vdisks.
There are no definite values, as each proper value depends on:
1) Size of the vdisk;
2) IOPS on the vdisk or storage device;
3) Performance of the storage devices and network (backups last longer on slow devices).
In case of Backup server issues, our developers recommend increasing the snap_reserve of that vdisk to 20% and, if the issues persist, to 30%.
Please note that the snapshot reserve space must be reserved at the time of snapshot creation in order to prevent any data corruption for backups. Having corrupted backups is worse than having none.
Finally, please note that increasing the snap_reserve space to high values may reserve more storage space than necessary for snapshots, resulting in wasted storage capacity, so it should be slowly increased only on vdisks with backup problems to find the proper value that resolves the backup issues.
If you would prefer to have the Support team perform the task for you, just create a ticket.