We are submitting a Request for Enhancement (RFE) to improve the integration between IBM's mmbackup utility for Scale and IBM Storage Protect (formerly TSM). This enhancement will address significant performance bottlenecks we've encountered when protecting large-scale Scale filesystems.
Background:
Our environment uses IBM Storage Protect to back up multiple Scale Filesets. Each fileset contains millions of files. The mmbackup tool does an excellent job of efficiently backing up this data.
However, a major challenge arises during the expiration process within IBM Storage Protect. By default, the Storage Protect server can only run one expiration process per filespace. A filespace is a logical representation of a filesystem backup. The default mmbackup configuration creates a single filespace for an entire Scale filesystem, even if that filesystem is massive and we have multiple filesets. This one-to-one relationship means that the expiration process for our large filesystems takes an extremely long time, often a week or more, which significantly impacts our operational efficiency.
Proposed Solution:
IBM Storage Protect offers a feature called virtualmountpoint to work around this limitation. This feature allows a single filesystem to be represented as multiple "virtual" filespaces within Storage Protect. Each virtual filespace can then be processed by its own expiration process, enabling parallel expiration and drastically reducing the time required.
We have already implemented a workaround to leverage this feature, but sometimes we got a mismatch and files ending up in our root fileset (filespace). Our current setup uses 10 dedicated nodes for backups. We have developed a custom script that runs before each backup to check if a virtual mount point exists for every fileset. If one is missing, our script updates all dsm.sys and then synchronizes the configuration across all 10 nodes.
This manual process is complex, time-consuming, and prone to error.
Request for Enhancement:
We request that a new feature be added to mmbackup to automate the creation and management of virtual mount points. Instead of requiring a separate, manual process, we would like mmbackup to be able to accept a parameter that tells it to automatically create a virtualmountpoint for each individual fileset within the filesystem.
This would simplify our backup administration, eliminate the need for our custom workaround scripts, and allow us to fully leverage the parallel processing capabilities of IBM Storage Protect for expiration and other maintenance tasks. This would be a significant quality-of-life improvement for administrators and would make mmbackup a more robust and scalable solution for managing massive GPFS filesystems.
we also run into this limitation. The idea to manage a list of virtual mountpoints was dropped because it is too much work to maintain it.