Tried retrying database dumps while running rsnapshot_backup, but gave up on this idea as research revealed flaws. For example, if repeating a file sync is appropriate because it continues a previously started process, then repeating heavy dumps may not be appropriate because it starts from the beginning every time and can delay the process as a whole.
Fixed the search for issues during time tracking from accounting: now searches for issues only in the header, raises an exception if several issues are found, etc.
We began to rework the descriptions of IP addresses in old client repositories from Jinja arrays to a neater yaml/sls format.
Fixed the error detection logic of the backup pipeline: if no check is performed, an error should be present, but not in the case when there are no backup descriptions.
Fixed mysql_replica_checker bug on old systems.
The ssh_keys state was not added to top sls in the salt repository template – fixed.