We're updating the issue view to help you get more done. 

Delay start of job dispatching on startup

Description

There are some cases where job dispatching too close to startup can cause operation failures.

One situation is where admin node is behind a proxy (a common configuration), so the admin node itself is on http://server.domain:8080/ but the proxy public URL is http://server.domain/

If Opencast restarts, the proxy (e.g. apache) can mark the service as offline and requests get a service-not-available 503 error for a short time even after the service has come back, because the proxy has a timeout, as described for the try parameter here:

https://httpd.apache.org/docs/current/mod/mod_proxy.html#proxypass

So Opencast checks for available services on http://server.domain/ and gets back nothing available even though it's up.

So it is safer to add a delay before starting to dispatch jobs on startup. This is most significant for scenarios where Opencast admin node is restarted while workflows are running (usually for unplanned reasons), and avoids workflows failing because no services are available.

Steps to reproduce

None

Status

Assignee

Unassigned

Reporter

Stephen Marquard

Criticality

None

Tags (folksonomy)

None

Components

Fix versions

Affects versions

3.3
4.0

Priority

Major