Description of problem: dstat not showing md devices Version-Release number of selected component (if applicable): pcp-dstat 5.0.2 How reproducible: dstat -D md0 Steps to Reproduce: 1. 2. 3. Actual results: You did not select any stats, using -cdngy by default. ----total-usage---- -net/total- ---paging-- ---system-- usr sys idl wai stl|| recv send| in out | int csw 0 0 100 0 0|2303 297 | 0 0 | 422 601 0 0 100 0 0|3072 268 | 0 0 | 246 252 0 0 100 0 0|2428 236 | 0 0 | 294 238 ^ Expected results: You did not select any stats, using -cdngy by default. ----total-usage---- --dsk/md0-- -net/total- ---paging-- ---system-- usr sys idl wai stl| read writ| recv send| in out | int csw 0 0 100 0 0| 0 0 |1868 640 | 0 0 | 191 194 0 0 100 0 0| 0 0 |1586 753 | 0 0 | 500 678 0 0 100 0 0| 0 0 |2833 268 | 0 0 | 242 230 Additional info: With version Dstat 0.7.2 this used to work.
Hmm, looking into it, it turns out it may be difficult to figure out the best path to take here. Old dstat used the following regex on device names to decide what to classify as a 'disk': ^([hsv]d[a-z]+\d+|cciss/c\d+d\d+p\d+|dm-\d+|md\d+|mmcblk\d+p\d0|VxVM\d+)$ As reported here, this definitely includes md and dm devices as disks. Somewhat counter intuitively, it excludes actual disks (!) as it considers only partitions as disks in the first part of the regex (this is a bug, partitioning is optional). One worthy goal here (and I believe this might be why partitions were favoured over disks originally) is to avoid double-accounting - IOW, summing the I/O to both a partition and the device the partitions sits on would make the 'total' column misleading. In the metrics namespace used by PCP and now dstat, we have 'disk' metrics in separate trees: $ pminfo disk.dev (raw disks - this is what dstat nowadays uses, but previously ignored in favour of partitions+DM+MD) $ pminfo disk.partitions disk.md disk.dm (the other classes of device, that sit atop disks) It turns out its not trivial to form a metric which is the union of all four sets, and its not clear we should do that anyway due to the double accounting issue. We could be bug-for-bug compatible with the original dstat, and use the union of partitions+DM+MD devices as 'disks' ... but that seems to me to be not ideal either (DM and MD devices are being built out of disks and/or partitions). I think the safest and best approach here will be to add dstat command line options for each class of device. That way, one knows exactly what they're looking at and we have the best chance of correctly accounting devices in the 'total' columns. Another smaller advantage of this approach of using disk.dev.* devices for --disk reporting is that this option will (continue to) work on platforms other than Linux. So to summarise, I'm proposing we leave --disk reporting for actual disks, as it does now, and add plugins/command line options for --part/--partitions, --dm/--device-mapper and --md/--multi-device driver devices. The 'total' column is then very well-defined for each case.
commit a88e8cef5f21e1d95b322fbefe8b7e5461ff4cb5 Author: Nathan Scott <nathans> Date: Wed Jan 29 16:49:07 2020 +1100 pcp-dstat: add options for md, dm and disk partition reporting Keep the --disk/-D options for (actual, physical) disks, and add new plugins and options for --dm, --md and --partition based reporting. This allows us to report these devices while continuing to report a reasonable --disk 'total' column. Resolves Fedora BZ #1794273
FEDORA-2020-1948419b2e has been submitted as an update to Fedora 31. https://bodhi.fedoraproject.org/updates/FEDORA-2020-1948419b2e
pcp-5.0.3-1.fc31 has been pushed to the Fedora 31 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2020-1948419b2e
pcp-5.0.3-1.fc32 has been pushed to the Fedora 32 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2020-7ac95d60db
FEDORA-2020-a4236a5dcf has been submitted as an update to Fedora 31. https://bodhi.fedoraproject.org/updates/FEDORA-2020-a4236a5dcf
pcp-5.0.3-2.fc31 has been pushed to the Fedora 31 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2020-a4236a5dcf
pcp-5.0.3-2.fc31 has been pushed to the Fedora 31 stable repository. If problems still persist, please make note of it in this bug report.
pcp-5.0.3-1.fc32 has been pushed to the Fedora 32 stable repository. If problems still persist, please make note of it in this bug report.