Hi Kristoffer, No, this isn't a known or normal issue. What version of Phoenix and HBase are you using? What does your DELETE statement look like? Perf-wise, it's best if you can: * Batch deletes (any mutation, really) to cut down on RPC traffic. The phoenix.mutate.batchSize config parameter controls that and defaults to 1000. * Run as few DELETE statements as possible. For example, if you're deleting by PK, then form a single DELETE statement with an IN clause in the WHERE clause. * Set auto commit on for your connection: conn.setAutoCommit(true) assuming you don't have to issue a DELETE statement for each row separately. * If you do have to issue separate DELETE statement for each row, then keep auto commit as false and instead collect up batches of 1000 before issuing a commit.
If you're doing all that, and it's still slow, check your logs for exceptions and perhaps try issuing a delete from the HBase shell against your table as a test. Thanks, James On Fri, Mar 28, 2014 at 8:46 AM, Kristoffer Sjögren <[email protected]>wrote: > Hi > > I was doing some data migration today on phoenix and noticed that delete > ... where statements are _really_ slow (around 1 second per row). > > Is this a known issue? Our installation works fine for inserts and selects. > > Cheers, > -Kristoffer >
