commits/26394

Change 26394

tom_tyler (C. Thomas Tyler)
C. Thomas Tyler committed this change 5 years ago into / under Review 26175
View Review
Download .zip
Provide snapshot hook for daily_checkpoint.sh.

Perhaps the most common SDP customization is a tweak to integrate
the SDP checkpoint mechanism with site-local capability of the
underlying hardware, so that the snapshot is created at the ideal
point in time, right after the metadata checkpoint is created.

The intent of this change is to reduce the need for such
customization by making a well defined hook in the daily checkpoint
process for calling a site-specific custom checkpoint script.

Key elements:
* Made an illustrative sample change to instance_vars.template,
to show setting a SNAPSHOT_SCRIPT variable defined only if
running on the master. This logic would be customized to meet
local needs, e.g. perhaps checking hostname if snapshots can
only be run on certiain machines.  The custom logic would live
in the /p4/common/config/p4_N.vars file.  As a for-example,
The sample logic shows a commented-out call to an Amazon Web
Services (AWS) Elastic Block Store (EBS) snapshot. (This could
just as easily reference a NetApp filer snapshot for an on-prem
installation.)
* A change to dump_checkpoint() in backup_functions.sh to
call the snapshot script defined by the $SNAPSHOT_SCRIPT variable
if defined.

As coded, a failure of either the checkpoint or the snapshot
will cause the script to report a failure. However, a failure
of the checkpoint will not prevent the snapshot from being
attempted.

Possible future tweaks: Consider adding a LIMIT_ONE_DAILY_SNAPSHOT
setting (akin to the LIMIT_ONE_DAILY_CHECKPOINT setting in the
Windows SDP).
SDP-436
Add Job
2 edited 0 added 0 deleted
Tip: Use n and p to cycle through the changes.
rwillyoung (rwillyoung)
rwillyoung commented 5 years ago

Hey Tom! Commentary from the peanut gallery!
Any chance of shooting an email to your partners at Netapp about integrating a call to netapp snapcenter? Not expecting perforce to actually add that but I'd hope/imagine that with you laying the ground work to support "wait for outside snapshot tool" that they could get on the bandwagon.
Reply
tom_tyler (C. Thomas Tyler)
C. Thomas Tyler commented 5 years ago

Good idea!

As a customization in several customer environments I've done it such that the Helix Core server does an SSH call to the NetApp filer, with a dependency on having SSH keys enabled so that the 'perforce' OS user on the Helix Core server machine can reach out to filer and run the command to kick off a snapshot. Add in some spice to give the snapshot a decent name. That works if you want the backups driven by the Helix Core server. An alternative (which wouldn't need this change) is to drop the crontabs, and have the Helix Core server's daily_checkpoint.sh called from the NetApp.

Similarly, there's an AWS sample already included in the SDP (not yet documented), which works with DBS snapshots.

I'd say this sort of local backup integration is the most common customization of the SDP, and I wanted that change to be more of a configuration rather than customization thing.
Reply