Bug: Deadlock encountered about surrealdb

[triracle97]

I encountered it again, the issue still persists, just harder to reproduce

[phughk]

Heya @triracle97 thanks for raising this. The live queries are going through a significant design change that will reduce the chance of this deadlock happening significantly. We are currently seeing another perf issue with 1.4.x in #3906 - it may be loosely related or at least correlated. Will have a look once 3906 is resolved.

[triracle97]

Thanks for replying @phughk!
Is there any side effect if I comment that line to prevent deadlock temporarily, I don't need to reuse live query and I can kill and recreate live query everytime I create new connection

[phughk]

Unfortunately, that isn't possible - the lock is required to check which connection to send live query notifications to.

[triracle97]

Update for this, since updated to 1.5.0, I haven't encountered this and can't reproduce it with above method

[phughk]

Thanks for the update and info - we will try to tackle this in. The solution would be to have channels with messages and a single handler. It's a bit of work and we are very focused on cloud and 2.0 at the moment. But this is on our radar. Hopefully retrying can be a way around this