Nova Scheduler warning "Host has more disk space than database expected" on launching instances

asked 2014-12-19 19:50:55 -0500

vahidh gravatar image

Hi there,

I'm running some experiments in which through a script I launch a number of instances one after the other (nova boot) and then remove them all (nova delete). And I repeat the scenario above a number of times.

The script runs fine for the first few rounds and then instance launches all fail with warnings like this in scheduler.log:

WARNING nova.scheduler.host_manager [req-30c6f9e6-54db-495c-a5ee-bc679179dcc8 f68eb94377af4ec69cded8d8e371c63a 800a856ea54041c7839585d10298ad05] Host has more disk space than database expected (815gb > 814gb)

Once this issue starts compute.log no longer receives any log messages without registering any error messages. Even though at each round OpenStack services including Nova are restarted. No other error messages are found in Nova logs.

Once the issue starts it continues. But usually a controller reboot brings things back to normal.

Any idea what could be the issue here?

Thanks.

edit retag flag offensive close merge delete

Comments

That's a great error. It might be a bug - why not link a copy of your script and lodge a bug at http://bugs.launchpad.net/nova/ tagged "ops"

fifieldt gravatar imagefifieldt ( 2015-01-09 03:52:10 -0500 )edit

I'm also experiencing this issue. It seems that the discussion around this has died. Is there a way to get the database to re-read disk space?

khrystoph gravatar imagekhrystoph ( 2015-05-20 00:10:01 -0500 )edit

I've the same error. I have installed RDO Kilo in CentOS 7.

sacha-m gravatar imagesacha-m ( 2015-06-17 21:55:34 -0500 )edit

I've the same error, Kilo in Ubuntu

hobby tan gravatar imagehobby tan ( 2016-10-17 21:50:37 -0500 )edit