Increasing RT2 timeouts [WAS: Repotool2 times out on harvest of MCOs]

[tomwrobel]

Many contributor objects (MCOs) where there are more than N contributors are failing harvest. We don't know what N is, only that objects in the thousands certainly fail.

Repotool2 allows 100 seconds for a Sword2 response. Therefore we have to respond within that time.

Sword2 has been updated (currently only on QA) to truncate the number of contributors at 100, and to put an et_al name element for the authors role. This should speed up the read, even if it means that the number of contributors is truncated.

Related commits:

• Willow sword [feature/ora_customizations 3bc9745] Add maximum contributors in response
• Ansible.ora4 [TW-add-max-contributors-config-option 706b736] Configure Sword2 to return N contributors only

[tomwrobel]

ancilliary question: does Repotool2 pick up the last item in an OAI-PMH list?

[tomwrobel]

Notes: we need to increase the timeout value in Repotool2, there is no easy way around this

[tomwrobel]

Notes: there is an intended fix for this by increasing timeouts in 5.20 that should be backported to 5.18. We will revisit this issue once that is deployed

[mrdsaunders]

From Andrew Bennett:
In 5.20 improvements to the crosswalk engine will solve the issue. For a paper with 2,800 authors:
16,234ms when limited to 25 authors
28,498ms when limited to 50 authors
48,802ms when limited to 100 authors
After 5.20 all 2,800 authors will be processed in 314ms.

For 5.18/5.19 (following patch) we can limit the number of authors via the crosswalk:
<xwalk:parameter name="person-list-entry-read-limit" value="50" />

I will test (once on 5.19) and then update the crosswalk.

[tomwrobel]

@tomwrobel to note before we talk about this ticket tomorrow, Jason has explicitly requested that the person list can be limited BUT

• we need all depositor and impersonator information
• we need all known Oxford Authors