Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Overcommit disk resource

As far as I know, disk resource overcommit is configured by the parameter disk_allocation_ratio.

By default, this parameter is set to 1.0, which means disk resource overcommit is not allowed. I tried to launch an instance which requires 80G disk space, which none of the compute nodes have sufficient space available. However, the instance still launched succesfully!

Here is the nova log:

2015-01-15 12:39:19.279 7145 DEBUG nova.scheduler.filter_scheduler [req-52b8c668-d33a-459d-b1cb-2849de2ac7a8 1ec3f91078d844b4ba20d5b30ec3f49c 226f89cdecef4f05a6931a4215b8a696] Filtered [(pct93, pct93) ram:43428 disk:-43008 io_ops:0 instances:1, (pct94, pct94) ram:47688 disk:-4096 io_ops:0 instances:1] _schedule /usr/lib/python2.6/site-packages/nova/scheduler/filter_scheduler.py:332
2015-01-15 12:39:19.280 7145 DEBUG nova.scheduler.filter_scheduler [req-52b8c668-d33a-459d-b1cb-2849de2ac7a8 1ec3f91078d844b4ba20d5b30ec3f49c 226f89cdecef4f05a6931a4215b8a696] Weighed [WeighedHost [host: pct94, weight: 1.0], WeighedHost [host: pct93, weight: 0.91066935078]] _schedule /usr/lib/python2.6/site-packages/nova/scheduler/filter_scheduler.py:337
2015-01-15 12:39:19.281 7145 INFO nova.scheduler.filter_scheduler [req-52b8c668-d33a-459d-b1cb-2849de2ac7a8 1ec3f91078d844b4ba20d5b30ec3f49c 226f89cdecef4f05a6931a4215b8a696] Choosing host WeighedHost [host: pct94, weight: 1.0] for instance 065aea35-4249-48de-9247-23591103f5a7

As shown in the logs, I have 2 compute nodes and both "disk" values are negative, which should mean there is no sufficient disk space available, why the host "pct94" is still chosen?

Thanks for anyone helping in advance.