Bug#1040415: bullseye-pu: package pacemaker/2.1.5-1+deb12u1
"Adam D. Barratt" writes: > Assuming that all of the changes are in unstable (or not required > there), please go ahead. Hi Adam, All three patches are cherry-picks (although one required manual backporting) from 2.1.6, which is already in testing. Source-only upload done, hope it's all right. -- Thanks, Feri.
Bug#1040415: bullseye-pu: package pacemaker/2.1.5-1+deb12u1
Control: tags -1 + confirmed On Sun, 2023-07-09 at 23:37 +0200, Ferenc Wágner wrote: > Control: tag -1 - confirmed > > Jonathan Wiltshire writes: > > > On Wed, Jul 05, 2023 at 07:14:09PM +0200, Ferenc Wágner wrote: > > > > > Shortly after the release of bookworm we got a report that > > > Pacemaker > > > regressed in certain migration scenarios when compared to the > > > bullseye > > > version. Upstream identified the cause (a bug already fixed in > > > 2.1.6), > > > and after backporting the fix the submitter acknowledged that > > > they can't > > > reproduce the bug anymore with the proposed packages. > > > https://bugs.clusterlabs.org/show_bug.cgi?id=5521 > > > Pacemaker package bug opened after discussion on the mailing > > > list: > > > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1040165 > > > > Please go ahead, and bear in mind the upload window closes next > > weekend. > > Thanks, Jonathan! Does it mean that I have to upload before 15th of > July? > If you want to be sure of the update making it into 12.1, yes. > On the other hand, meanwhile upstream notified me that to fully fix > this > bug I need to backport one more patch, which in turn required > including > a third one. So the debdiff grew a little, please reconfirm the > upload: > Assuming that all of the changes are in unstable (or not required there), please go ahead. Regards, Adam
Bug#1040415: bullseye-pu: package pacemaker/2.1.5-1+deb12u1
Control: tag -1 - confirmed
Jonathan Wiltshire writes:
> On Wed, Jul 05, 2023 at 07:14:09PM +0200, Ferenc Wágner wrote:
>
>> Shortly after the release of bookworm we got a report that Pacemaker
>> regressed in certain migration scenarios when compared to the bullseye
>> version. Upstream identified the cause (a bug already fixed in 2.1.6),
>> and after backporting the fix the submitter acknowledged that they can't
>> reproduce the bug anymore with the proposed packages.
>> https://bugs.clusterlabs.org/show_bug.cgi?id=5521
>> Pacemaker package bug opened after discussion on the mailing list:
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1040165
>
> Please go ahead, and bear in mind the upload window closes next weekend.
Thanks, Jonathan! Does it mean that I have to upload before 15th of July?
On the other hand, meanwhile upstream notified me that to fully fix this
bug I need to backport one more patch, which in turn required including
a third one. So the debdiff grew a little, please reconfirm the upload:
$ debdiff pacemaker_2.1.5-1.dsc pacemaker_2.1.5-1+deb12u1.dsc
dpkg-source: warning: extracting unsigned source package
(/home/wferi/ha/pacemaker/pacemaker_2.1.5-1.dsc)
diff -Nru pacemaker-2.1.5/debian/changelog pacemaker-2.1.5/debian/changelog
--- pacemaker-2.1.5/debian/changelog2023-01-22 16:38:34.0 +0100
+++ pacemaker-2.1.5/debian/changelog2023-07-09 23:10:45.0 +0200
@@ -1,3 +1,17 @@
+pacemaker (2.1.5-1+deb12u1) bookworm; urgency=medium
+
+ * [0c22be8] New patches fixing migration regression.
+Backport of https://github.com/ClusterLabs/pacemaker/pull/3020/ to
+Pacemaker 2.1.5 (without the CTS changes, which we don't ship):
+5754a2af9 Refactor: scheduler: improve xpath efficiency when unpacking
+3f6f524f1 Low: scheduler: unknown_on_node() should ignore pending actions
+ad9fd9548 Fix: scheduler: handle cleaned migrate_from history correctly
+The starting refactor is required by the other two patches, but the
+third patch still needed backporting.
+Thanks to Ken Gaillot (Closes: #1040165)
+
+ -- Ferenc Wágner Sun, 09 Jul 2023 23:10:45 +0200
+
pacemaker (2.1.5-1) unstable; urgency=medium
* [5792d59] Work around lazy loading of GitHub release pages in watch file
diff -Nru pacemaker-2.1.5/debian/gbp.conf pacemaker-2.1.5/debian/gbp.conf
--- pacemaker-2.1.5/debian/gbp.conf 2023-01-22 13:10:39.0 +0100
+++ pacemaker-2.1.5/debian/gbp.conf 2023-07-09 22:33:06.0 +0200
@@ -1,5 +1,5 @@
[DEFAULT]
-debian-branch = debian/master
+debian-branch = debian/bookworm
upstream-branch = upstream/latest
[import-orig]
diff -Nru
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
---
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
1970-01-01 01:00:00.0 +0100
+++
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
2023-07-09 23:07:30.0 +0200
@@ -0,0 +1,30 @@
+From: Ken Gaillot
+Date: Wed, 1 Feb 2023 17:12:13 -0600
+Subject: Fix: scheduler: handle cleaned migrate_from history correctly
+
+Fixes T623
+---
+ lib/pengine/unpack.c | 10 ++
+ 1 file changed, 10 insertions(+)
+
+diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
+index e9fcae1..99a2dc4 100644
+--- a/lib/pengine/unpack.c
b/lib/pengine/unpack.c
+@@ -2937,6 +2937,16 @@ unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t
*node, xmlNode *xml_op,
+ }
+
+ } else { // Pending, or complete but erased
++
++/* If there is no history at all for the resource on an online
target, then
++ * it was likely cleaned. Just return, and we'll schedule a probe.
Once we
++ * have the probe result, it will be reflected in target_newer_state.
++ */
++if ((target_node != NULL) && target_node->details->online
++&& unknown_on_node(rsc, target)) {
++return;
++}
++
+ /* If the resource has newer state on the target, this migrate_to no
+ * longer matters for the target.
+ */
diff -Nru
pacemaker-2.1.5/debian/patches/Low-scheduler-unknown_on_node-should-ignore-pending-actio.patch
pacemaker-2.1.5/debian/patches/Low-scheduler-unknown_on_node-should-ignore-pending-actio.patch
---
pacemaker-2.1.5/debian/patches/Low-scheduler-unknown_on_node-should-ignore-pending-actio.patch
1970-01-01 01:00:00.0 +0100
+++
pacemaker-2.1.5/debian/patches/Low-scheduler-unknown_on_node-should-ignore-pending-actio.patch
2023-07-09 23:07:30.0 +0200
@@ -0,0 +1,80 @@
+From: Ken Gaillot
+Date: Thu, 2 Feb 2023 10:25:53 -0600
+Subject: Low: scheduler: unknown_on_node() should ignore pending actions
+
+Previously, unknown_on_node() looked for any lrm_rsc_op at all to decide
+whether a resource is known on
Bug#1040415: bullseye-pu: package pacemaker/2.1.5-1+deb12u1
Control: tag -1 confirmed On Wed, Jul 05, 2023 at 07:14:09PM +0200, Ferenc Wágner wrote: > [ Reason ] > Shortly after the release of bookworm we got a report that Pacemaker > regressed in certain migration scenarios when compared to the bullseye > version. Upstream identified the cause (a bug already fixed in 2.1.6), > and after backporting the fix the submitter acknowledged that they can't > reproduce the bug anymore with the proposed packages. > https://bugs.clusterlabs.org/show_bug.cgi?id=5521 > Pacemaker package bug opened after discussion on the mailing list: > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1040165 Please go ahead, and bear in mind the upload window closes next weekend. Thanks, -- Jonathan Wiltshire j...@debian.org Debian Developer http://people.debian.org/~jmw 4096R: 0xD3524C51 / 0A55 B7C5 1223 3942 86EC 74C3 5394 479D D352 4C51 ed25519/0x196418AAEB74C8A1: CA619D65A72A7BADFC96D280196418AAEB74C8A1
Bug#1040415: bullseye-pu: package pacemaker/2.1.5-1+deb12u1
Package: release.debian.org
Severity: normal
Tags: bookworm
User: release.debian@packages.debian.org
Usertags: pu
Dear Stable Release Team,
[ Reason ]
Shortly after the release of bookworm we got a report that Pacemaker
regressed in certain migration scenarios when compared to the bullseye
version. Upstream identified the cause (a bug already fixed in 2.1.6),
and after backporting the fix the submitter acknowledged that they can't
reproduce the bug anymore with the proposed packages.
https://bugs.clusterlabs.org/show_bug.cgi?id=5521
Pacemaker package bug opened after discussion on the mailing list:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1040165
[ Impact ]
Core HA functionality is impacted, there's no easy way to work around
the problem. Pacemaker 2.1.5-1 is unsuitable for big portion of its
intended applications.
[ Tests ]
The submitter tested and confirmed the fix.
[ Risks ]
The patch is small but the backport wasn't trivial due to extensive
refactorings meanwhile. I asked upstream to sanity-check it, but
haven't got a reply yet.
[ Checklist ]
[x] *all* changes are documented in the d/changelog
[x] I reviewed all changes and I approve them
[x] attach debdiff against the package in (old)stable
[x] the issue is verified as fixed in unstable
$ debdiff pacemaker_2.1.5-1.dsc pacemaker_2.1.5-1+deb12u1.dsc
diff -Nru pacemaker-2.1.5/debian/changelog pacemaker-2.1.5/debian/changelog
--- pacemaker-2.1.5/debian/changelog2023-01-22 16:38:34.0 +0100
+++ pacemaker-2.1.5/debian/changelog2023-07-02 21:39:59.0 +0200
@@ -1,3 +1,11 @@
+pacemaker (2.1.5-1+deb12u1) bookworm; urgency=medium
+
+ * [20411a8] New patch: Fix: scheduler: handle cleaned migrate_from history
+correctly.
+Thanks to Ken Gaillot (Closes: #1040165)
+
+ -- Ferenc Wágner Sun, 02 Jul 2023 21:39:59 +0200
+
pacemaker (2.1.5-1) unstable; urgency=medium
* [5792d59] Work around lazy loading of GitHub release pages in watch file
diff -Nru pacemaker-2.1.5/debian/gbp.conf pacemaker-2.1.5/debian/gbp.conf
--- pacemaker-2.1.5/debian/gbp.conf 2023-01-22 13:10:39.0 +0100
+++ pacemaker-2.1.5/debian/gbp.conf 2023-07-02 21:39:59.0 +0200
@@ -1,5 +1,5 @@
[DEFAULT]
-debian-branch = debian/master
+debian-branch = debian/bookworm
upstream-branch = upstream/latest
[import-orig]
diff -Nru
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
---
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
1970-01-01 01:00:00.0 +0100
+++
pacemaker-2.1.5/debian/patches/Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
2023-07-02 21:39:59.0 +0200
@@ -0,0 +1,30 @@
+From: Ken Gaillot
+Date: Wed, 1 Feb 2023 17:12:13 -0600
+Subject: Fix: scheduler: handle cleaned migrate_from history correctly
+
+Fixes T623
+---
+ lib/pengine/unpack.c | 10 ++
+ 1 file changed, 10 insertions(+)
+
+diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
+index 5fcba3b..abfd06f 100644
+--- a/lib/pengine/unpack.c
b/lib/pengine/unpack.c
+@@ -2920,6 +2920,16 @@ unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t
*node, xmlNode *xml_op,
+ }
+
+ } else { // Pending, or complete but erased
++
++/* If there is no history at all for the resource on an online
target, then
++ * it was likely cleaned. Just return, and we'll schedule a probe.
Once we
++ * have the probe result, it will be reflected in target_newer_state.
++ */
++if ((target_node != NULL) && target_node->details->online
++&& unknown_on_node(rsc->id, target, data_set)) {
++return;
++}
++
+ /* If the resource has newer state on the target, this migrate_to no
+ * longer matters for the target.
+ */
diff -Nru pacemaker-2.1.5/debian/patches/series
pacemaker-2.1.5/debian/patches/series
--- pacemaker-2.1.5/debian/patches/series 2023-01-22 13:31:42.0
+0100
+++ pacemaker-2.1.5/debian/patches/series 2023-07-02 21:39:59.0
+0200
@@ -5,3 +5,4 @@
Shipping-the-CTS-is-not-useful.patch
Always-run-Inkscape-under-the-C.UTF-8-locale.patch
Fix-typos-resouce-resource.patch
+Fix-scheduler-handle-cleaned-migrate_from-history-correct.patch
diff -Nru pacemaker-2.1.5/debian/salsa-ci.yml
pacemaker-2.1.5/debian/salsa-ci.yml
--- pacemaker-2.1.5/debian/salsa-ci.yml 2023-01-22 13:10:39.0 +0100
+++ pacemaker-2.1.5/debian/salsa-ci.yml 2023-07-02 21:39:59.0 +0200
@@ -5,6 +5,7 @@
variables:
SALSA_CI_REPROTEST_ENABLE_DIFFOSCOPE: 1
+ RELEASE: bookworm
autopkgtest:
extends: .test-autopkgtest
--
Thanks,
Feri.
