On 11/08/2017 09:51 AM, Wolfgang Bumiller wrote:
> Turns out some of the HA code cannot properly deal with
> cfs_read_file() die()ing, so make this behavior (introduced
> in 7bac9ca573ad ("cluster: improve error handling when
> reading files")) optional and off-by-default for now.
>
> This can later be re-enabled once all affected callers have
> been dealt with.
>
The main trouble maker is our ha manager as it is the only one
which not only simply dies itself, as e.g., an API call, but may
bring the whole node down, if this happens.
If something other dies we should not get into real problems,
so, IMHO, it would be cleaner to let this stay and address it
in the ha-manager, as my proposed series does.
_______________________________________________
pve-devel mailing list
pve-devel@pve.proxmox.com
https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
