Bulk actions fail altogether (nothing happens) if too many machines are selected

[Andres Rodriguez]

When selecting 12 nodes to perform a bulk action in them, nothing really happens. The UI does not provide any feedback and the nodes don't seem to start commissioning.

However, when selecting only 5, this works just fine.

It seems that when many nodes are used for a bulk action, the bulk action might be failing altogether.

[Andres Rodriguez (andreserl)]

There's nothing at the logs that give us an idea of what might be going wrong!

[Raphaël Badin (rvb)]

As discussed during the stand-up today, this is probably related to the fact that MAAS uses a thread pool (of 10) to deal with async calls.

[Raphaël Badin (rvb)]

When changing the number of threads, things definitely improve… but this doesn't explain what the core problem is: even with a limited number of threads things should just be slow but work: it seems that when more actions than the number of threads are started, things deadlock. When commissioning 11 nodes (with the default thread poll of 10 threads), things would deadlock and the last query in the DB's log is UPDATE "metadataserver_nodeuserdata" SET "node_id" = 4, "data" = '<blah>'; after this things came to a halt