<div dir="ltr"><div dir="ltr">On Mon, Feb 16, 2026 at 3:52 AM Alexander Semke <<a href="mailto:alexander.semke@web.de">alexander.semke@web.de</a>> wrote:</div><div class="gmail_quote gmail_quote_container"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 13/02/26 12:37, Ben Cooksley wrote:<br>
> [...]<br>
> Resource utilisation wise, i've not looked into whether there has been <br>
> a significant bump in the number of jobs, but over the past year some <br>
> additional CD support has been added so that indicates some trouble there.<br>
> System utilisation stats since the start of February are rather telling<br>
><br>
>           full_path           |    time_used     | job_count<br>
> ------------------------------+------------------+-----------<br>
> graphics/krita               | 228:03:25.300251 |       451<br>
> plasma/kwin                  | 164:34:16.119352 |      2032<br>
> plasma/plasma-workspace      | 136:03:27.534157 |      1044<br>
> multimedia/kdenlive          | 113:13:39.087112 |       566<br>
> network/ruqola               | 89:02:24.704107  |       369<br>
> pim/messagelib               | 83:58:38.936333  |       648<br>
> graphics/drawy               | 77:38:55.987716  |      2101<br>
> network/neochat              | 76:37:12.290371  |      1069<br>
> plasma/plasma-desktop        | 62:27:04.468738  |       619<br>
> utilities/kate               | 59:05:46.517227  |       438<br>
> office/kmymoney              | 56:58:08.349153  |       241<br>
> education/labplot            | 55:19:06.879103  |       423<br>
> frameworks/kio               | 43:04:55.33379   |       744<br>
> kde-linux/kde-linux-packages | 39:40:15.332536  |        63<br>
> libraries/ktextaddons        | 31:46:16.168865  |       356<br>
> office/calligra              | 31:33:35.74806   |        62<br>
> kdevelop/kdevelop            | 30:57:38.655126  |       113<br>
> education/kstars             | 30:04:42.328143  |        71<br>
> sysadmin/craft-ci            | 28:36:13.927477  |        53<br>
> bjordan/kdenlive             | 24:13:20.564568  |       191<br>
><br>
><br>
> Contrast that with the full month of September 2025:<br>
><br>
>            full_path           |    time_used     | job_count<br>
> -------------------------------+------------------+-----------<br>
> graphics/krita                | 382:38:14.237481 |       882<br>
> plasma/kwin                   | 365:27:01.96681  |      3531<br>
> multimedia/kdenlive           | 294:59:01.874773 |      1299<br>
> network/neochat               | 233:26:31.601019 |      2298<br>
> packaging/flatpak-kde-runtime | 225:01:38.542109 |       269<br>
> network/ruqola                | 219:56:34.415103 |       946<br>
> plasma/plasma-workspace       | 194:15:09.699541 |      2014<br>
> sysadmin/ci-management        | 118:10:28.754041 |       233<br>
> libraries/ktextaddons         | 105:25:58.724964 |      1084<br>
> office/kmymoney               | 101:50:27.612055 |       572<br>
> plasma/plasma-desktop         | 101:25:05.185953 |      1148<br>
> kde-linux/kde-linux-packages  | 101:17:06.284095 |       144<br>
> education/rkward              | 84:32:07.118316  |      1033<br>
> kdevelop/kdevelop             | 84:04:58.804139  |       270<br>
> utilities/kate                | 83:49:15.561603  |       663<br>
> frameworks/kio                | 76:46:04.377597  |      1032<br>
> graphics/okular               | 74:58:19.8886    |       628<br>
> pim/akonadi                   | 62:23:56.252831  |       531<br>
> pim/itinerary                 | 60:05:52.269941  |       578<br>
> network/kaidan                | 56:37:44.446695  |      1773<br>
<br>
<br>
Thanks Ben, this is an interesting information. Would it be feasible for <br>
you to send such a report to the mailing list on the monthly base? Or <br>
even add this as the output for every merge request on gitlab? I think <br>
teams can probably easily improve in the various areas, at least in some <br>
of them, and optimize the build and the tests to reduce the overall <br>
runtime. Providing more transparency into the consumption of resources <br>
on CI per project, per job and per test can support such optimizations <br>
and will also help to raise the awareness for the impact of certain <br>
changes on CI.<br></blockquote><div><br></div><div>It has to be manually put together based on queries from the database, so can't be put on every MR.</div><div>It can be run periodically though.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
For me, waiting for CI is really annoying and also disrupting personal <br>
workflows sometimes (waiting today for 3h already in <br>
<a href="https://invent.kde.org/education/labplot/-/merge_requests/845" rel="noreferrer" target="_blank">https://invent.kde.org/education/labplot/-/merge_requests/845</a>, gave up <br>
and switched to a different branch already and will need to go back in <br>
case of errors). For LabPlot we did already multiple optimizations in</blockquote><div><br></div><div>Looks like we dropped 2 runners again (node2 and node3) which i've now revived.</div><div>This hasn't been helped by a number of large projects doing quite a bit of activity today (which is unusual for a Sunday)</div><div><br></div><div><span style="font-family:monospace"><span style="color:rgb(0,0,0)">             full_path              |    time_used    | job_count  </span><br>------------------------------------+-----------------+-----------
<br> plasma/plasma-desktop              | 07:25:58.851394 |        44
<br> graphics/krita                     | 07:16:57.78504  |        12
<br> multimedia/kdenlive                | 05:17:33.323612 |        30
<br> office/kmymoney                    | 03:54:42.976902 |        16
<br> pehg/kaidan                        | 03:16:03.91073  |        30
<br> snap-packaging/kf6-core-sdk        | 03:15:04.291499 |         3
<br> system/dolphin                     | 03:10:27.969557 |        48
<br> utilities/ark                      | 03:04:04.142    |        33
<br> network/neochat                    | 02:59:31.366648 |        85
<br> education/labplot                  | 02:50:22.827995 |        23
<br> libraries/ktextaddons              | 02:44:16.520082 |        35
<br> pim/messagelib                     | 02:17:45.340389 |        14
<br> pim/itinerary                      | 02:14:12.165157 |         8
<br> cwo/plasma-desktop                 | 02:09:26.453516 |         7
<br> deeptangshusaha/kdeconnect-android | 01:42:42.818736 |         9
<br> office/skrooge                     | 01:35:08.920711 |         9
<br> qazcetelic/kdeconnect-android      | 01:13:10.125977 |         6
<br> graphics/drawy                     | 01:11:45.894211 |        44
<br> lito/kdeconnect-kde                | 01:06:17.659945 |        24
<br> graphics/digikam                   | 01:05:11.594924 |        13<br>
<br></span></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> <br>
the past - not because of the long pending time but just because the <br>
runtime itself was painful and we plan to do another iteration and to <br>
invest more here soon. The problem with the pending time come on top now <br>
and I think teams should check what can be improved here in general <br>
without putting the focus on more hardware only.<br></blockquote><div><br></div><div>Optimisations, especially when they are things that should be done anyway - like removing tests that are broken (timeout, fail) or which don't actually test anything of meaning - are always useful.</div><div><br></div><div>I do believe in this instance there is a hardware element to it - even if it is to simply improve reliability. </div><div><br></div><div>i've noticed that crashes tend to happen most frequently during the launch stage as well, so i'm considering adding locking to vm-runner so only one VM can be starting at any given moment - there is a possibility we are triggering a bug in KVM itself (or elsewhere in the kernel as shipped by Debian). Given the reliability has been declining since the initial introduction of VM based CI i'm not 100% convinced this is the case though.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
<br>
-- <br>
<br>
Alexander<br>
<br></blockquote><div><br></div><div>Cheers,</div><div>Ben</div></div></div>