Running ipdevpoll in multiprocess mode will delay resolving of snmpAgentDown alerts

[lunkwill42]

A new multiprocess mode was introduced for ipdevpoll in NAV 4.7, but the snmpcheck plugin/job keeps some state internal to the running process. Under the new multiprocess mode, it's more or less random which worker process will process each snmpcheck job, meaning the internal state will become unsynced between workers.

The net effect is that once an snmpAgentDown alert has been issued, it can only be resolved if the same worker process happens to process the snmpcheck job for the same device again later, or if a new worker process is spawned and runs the job (since a new process will get its state from the database).

For the user, the symptom is that it appears that NAV does not detect, in a timely manner (if at all) that snmpAgentDown alerts are resolved.

This issue was already fixed on master by 12da0c2 - which should be cherry-picked onto 4.7.x for the next bugfix release.