So there is something that we've found in 1.4 with regards to /update requests. Essentially DirectUpdateHandler2, when going through it's commit method has a lock on itself. During this lock, it will do a full warmup of the new searchers caches. What this means is that DirectUpdateHandler2 might be locked up for a while. During this time, no /update queries will execute. They will just hang and end up with a very long QTime. Best advice if this is a problem for you.... upgrade. At least in 4.5 this lock no longer exists. It is not known exactly in what version you can get around this behavior.
Have more questions? Submit a request