bihealth / sodar-server

SODAR: System for Omics Data Access and Retrieval
https://github.com/bihealth/sodar-server
MIT License
14 stars 3 forks source link

Timeout in landing_zone_move during PREPARING state with large zone #1872

Closed mikkonie closed 8 months ago

mikkonie commented 8 months ago

I've received a report of a large landing zone (>160k files) timing out in the PREPARING state when landing_zone_move is triggered with validate_only=False. The operation has failed with Could not receive server response which usually indicates an iRODS session timeout.

My initial assumption is that the state simply takes too long without iRODS activity and the session times out. This happened after 8 minutes in the case which was reported.

mikkonie commented 8 months ago

After some experimentation, it seems the use of get_subcoll_obj_paths() may be the root cause of this. See #1882.

mikkonie commented 8 months ago

After deploying the changes discussed above, the PREPARING state still takes an inordinate time, but we don't get an iRODS timeout anymore. Considering the issue closed, although further optimization is most likely in order.