| commit | 8d2f4dc8010dc6bb05a0ba8d53be0490c2f145e6 | |
| author | Marko Myllynen <myllynen@redhat.com> | |
| Tue, 26 Jul 2016 01:39:49 +0000 (26 11:39 +1000) | ||
| committer | Nathan Scott <nathans@redhat.com> | |
| Tue, 26 Jul 2016 01:39:49 +0000 (26 11:39 +1000) | ||
| tree | f60dd2ca4663ca46aa3239f9422731ce2f3d7e76 | treesnapshot (tar.gz zip) |
| parent | 48c1696f1b29097311f790a9bc56c90b10640ae4 | commitdiff |
libvirt pmda: support for backing chains
Given the recent (IRC) discussion around per-device metrics and the
importance of keeping PMIDs unchanged for metrics, I think it's best to
add support for block device backing chains now after all so that we
won't end up to a situation later where we've added per-device metrics
(one way or another) and find out that adding backing chain statistics
might cause some PMID turbulence.
The patch is pretty trivial, extra handling is warranted as otherwise
the total values would be completely meaningless (backingIndex) or even
misleading (name). Also minor cosmetic reordering of conf file handling.
Reorder the list of block metrics to match libvirt ordering.
Given the recent (IRC) discussion around per-device metrics and the
importance of keeping PMIDs unchanged for metrics, I think it's best to
add support for block device backing chains now after all so that we
won't end up to a situation later where we've added per-device metrics
(one way or another) and find out that adding backing chain statistics
might cause some PMID turbulence.
The patch is pretty trivial, extra handling is warranted as otherwise
the total values would be completely meaningless (backingIndex) or even
misleading (name). Also minor cosmetic reordering of conf file handling.
Reorder the list of block metrics to match libvirt ordering.