Description of problem: With default value of 8192K (8MB), a reasonable 20MBps upload means a new ticket (and the whole process around it) 2.5 times a second or so. This is causing a performance issue. For example, changing to 819200K (800MB) improved performance (on my o-s-t environment, on my laptop!) from 20MBps to 60MBps. This is probably a short term tweak to improve the upload performance (I think further changes are needed to understand what we can improve here).
*** Bug 1508779 has been marked as a duplicate of this bug. ***
Increased UploadImageChunkSizeKB to 100MB
Issue verified. Engine build: ovirt-engine-4.2.0-0.0.master.20171114111003.git7aa1b91.el7.centos.noarch VDSM: 4.20.7-1.gitc9cf1ee Running the same upload disk test on the same stand (laptop-> engine -> vdsm host) with Increased UploadImageChunkSizeKB to 100MB on engine + new SDK script from my laptop improve the performance by ~22%. This is persistence for any stand I run with. Whatever stand I use (physical/Nested) using ISCSI storage domain(XtreamIO) the best time I get to upload a 10G preallocated policy with raw format image takes at best 14min. Details on runs made : I ran on my regular nested (storage-ge-04 environment) with the new script took 14min(~22% better). ~14min which is better than the ~18min-21min we got without the script so there is a major improvement there. I ran on 4.1 physical stand a manual (via UI) upload of the same 10G image & it took ~19 min - so no real change here between nested & physical engine/VDSMhost. On a 4.2 physical stand with the newest code via UI it took 22 min. On a 4.2 physical stand with newest code + new script took 17 min , a 5min improvement(~22% better) Vs UI.
This bugzilla is included in oVirt 4.2.0 release, published on Dec 20th 2017. Since the problem described in this bug report should be resolved in oVirt 4.2.0 release, published on Dec 20th 2017, it has been closed with a resolution of CURRENT RELEASE. If the solution does not work for you, please open a new bug report.