During snapshot restore, if certain API commands are executed through CLI/RestAPI, these may cause deadlocks in the system.
In the documentation we instruct the user to activate maintenance prior to restoring a snapshot. This is not doable through the UI because once a system is in maintenance mode the user cannot run restore. In this case we should alert the user and share two options:
run restore through the UI and make sure they dont run any external commands during the restore.
(recommended) switch to maintenance mode and use the CLI/Rest to start the restore.
We should document this for now so users will be clear on what they should do. we will apply the controls in a later release.
Depending on what we do with snapshot improvements, this may become unnecessary anyway.
I’m a bit lost what approach do we want to proceed with?
Is it just documenting or we want to skip this one until we will be doing work around the snapshots?
I think we should document the snapshot and restore process as part of the testing phase. try to be as clear as possible so that customers will not make mistakes. list all the Dos and Don’ts.
In 5.2 we will improve the mechanism and ideally reduce the potential erros.
Sounds sensible to me.