Now we're talking... and the talk is that the main Galera recv loop was absent from the logs from 7:44... to 7:57. Apparently it was stuck somewhere. I can tell that it was stuck at least 3 events before it had to send the state exchange message, however I don't know where exactly... In any case, this was caused by some extraordinary conditions on the box.
Just in case: what was the transaction rate there during SST and how much RAM was there?
Now we're talking... and the talk is that the main Galera recv loop was absent from the logs from 7:44... to 7:57. Apparently it was stuck somewhere. I can tell that it was stuck at least 3 events before it had to send the state exchange message, however I don't know where exactly... In any case, this was caused by some extraordinary conditions on the box.
Just in case: what was the transaction rate there during SST and how much RAM was there?