0

Rethinkdb IO reaches 100% whenever there is a data upload. The load reaches near about 50. Is this a common phenomenon, or do we need to do some optimizations here?

Ahmed Ashour
  • 4,209
  • 10
  • 29
  • 46
hellodk
  • 266
  • 3
  • 9
  • Hi @hellodk, under what circumstances is this happening? Could you tell us a little bit more about what you are running RethinkDB on? How are you connecting to it and what the insert code looks like? – dalanmiller Sep 30 '15 at 17:36
  • hi @dalanmiller the insert statement used here is: rethinkdb import -f .csv --format csv --table --pkey --force The rethinkdb instance is running on a 14 GB dual core machine
     – hellodk Oct 02 '15 at 14:00

1 Answers1

1

RethinkDB uses a blocker pool to do IO. On Linux systems, each thread in this blocker pool contributes 1 to the load average while blocking on disk, so RethinkDB sometimes causes the system to report an extremely high load average even under normal load.

Using 100% of your disk throughput is a different story. If you're running an IO-heavy workload on a slow disk, especially on a rotating drive, then that's pretty reasonable, but it does mean that you might have scaling problems if you want to do more disk-intensive operations. If you start to have those scaling problems, probably the best solution would be to get a faster disk.

mlucy
  • 5,046
  • 1
  • 14
  • 20
  • we're using ssd in our case, plus the task is to add some around 1M rows. I believe this is not a very huge figure to get 100% IO, or is there something which I missed? – hellodk Oct 02 '15 at 14:04
  • When you say 100% IO, do you mean IOWait, or are you referring to a different statistic? 100% IOWait isn't a problem unless you're also seeing slow performance, it just means that the disk is your bottleneck. If you're seeing slow performance along with 100% IOWait, I'd check to make sure you have enough RAM and that RethinkDB isn't going into swap. – mlucy Oct 02 '15 at 23:19