Hi Serega, How iostat looks like? are you seeing high utilization (above 60% across the drives) or long await write times when its writing to the WAL? What else do you see in the RS logs?
cheers, esteban. -- Cloudera, Inc. On Wed, Apr 22, 2015 at 10:39 AM, Serega Sheypak <serega.shey...@gmail.com> wrote: > Hi, we have 10 nodes cluster running HBase 0.98 CDH 5.2.1 > Sometimes HBase stucks. > We have several apps constantly writing/reading data from it. Sometimes we > see that apps response time dramatically increases. It means that app > spends seconds to read/write from/to HBase. in 99% of time it takes 20ms. > > I suppose that compactions/major compactions could be the root cause. I see > that compactions start at the same time when we have problems with app. > Could it be so? > So HBase can't write to WAL because compactions consumes all IO and apps > stops to write data? >