VBoxManage bandwidthctl breaks on low values => fixed in SVN/next maintenance

[dimaqq]

Setting disk bandwidth to a really low value (e.g. 1K) appears to turn off the throttle entirely, and throttle can only be re-enabled by vm power cycle (reboot is not enough).

Test: echo 3 > /proc/sys/vm/drop_caches time dd if=/dev/sdb of=/dev/null bs=1M count=10

Setup: Cloned a VM Set up IO group with 10M limit Attached existing 2nd SCSI disk to group

Test sequence: set limit to 1M booted up vm test result 1.2 MB/s repeated test (to be sure) test result 1.2 MB/s

set limit to 1K test result 91.9 MB/s

set limit to 1M test result 156 MB/s repeated test test result 82.0 MB/s

rebooted vm test result 82.0 MB/s

shut down vm started vm test resut 1.2 MB/s

Summary: It doesn't matter if low limit was set at runtime or in advance Critical value is between 210K (broken) and 225K (works) VM shutdown is needed to recover throttling

[aeichner]

Thanks for the report, this will be fixed in the next maintenance release.

You will observe a somewhat higher throughput for really low values because VBox will process at least one I/O request from the guest per second, even if the request size is bigger than the limit to avoid stalling I/O completely.

You might also see more throughput for certain disk images because VBox limits only real reads/writes to the image, reading from an image with unallocated or compressed blocks will result in higher throughput in the guest.

[Frank Mehnert]

Fix is part of 5.1.4.