[pve-devel] [PATCH v3 guest-common 1/2] fix #1291: implement remove_vmid_from_backup_jobs

[Fabian Grünbichler]

On September 4, 2019 4:41 pm, Thomas Lamprecht wrote:
> On 01.07.19 15:43, Christian Ebner wrote:
>> remove_vmid_from_backup_jobs updates the vzdump.cron backup jobs,
>> excluding the given vmid.
>> 
>> Signed-off-by: Christian Ebner <c.ebner at proxmox.com>
>> ---
>>  PVE/VZDump/Plugin.pm | 46 ++++++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 46 insertions(+)
>> 
>> diff --git a/PVE/VZDump/Plugin.pm b/PVE/VZDump/Plugin.pm
>> index 9933ef6..f415242 100644
>> --- a/PVE/VZDump/Plugin.pm
>> +++ b/PVE/VZDump/Plugin.pm
>> @@ -7,6 +7,8 @@ use POSIX qw(strftime);
>>  
>>  use PVE::Tools;
>>  use PVE::SafeSyslog;
>> +use PVE::Cluster qw(cfs_read_file cfs_write_file cfs_lock_file);
>> +use PVE::API2::Backup;
>        ^^^^^^^^^^^^^^^^^
> above won't fly, that's a module from pve-manager and thus would create a cyclic
> build dependency.. And I'd rather reduce than increase them ;)
> 
> The use is for the vzdump cron parser/writer which are cfs_registered in that
> file.. So either we move that out to a module here (or even higher up) or do
> something else, avoiding use of modules which are lower in the dependency chain.
> 

we want to use it from pve-container and qemu-server, so the only sane 
choice is to move the cfs_register part to guest-common IMHO. polluting 
pve-common or pve-cluster (or pve-access-control) is even worse, and 
there's nothing else that fits the bill that we can access from 
pve-container and qemu-server unless I am missing something..