Am 23.07.25 um 3:33 PM schrieb Fiona Ebner:
> Am 23.07.25 um 3:21 PM schrieb Fiona Ebner:
>> With the switch to '-blockdev', it is necessary to explicitly specify
>> the device ID for SCSI drives, see also [0]. Otherwise, the device
>> paths in '/dev/disk/by-id' would be missing. For IDE/SATA, the
On Wed, 23 Jul 2025 16:21:06 +0200, Stefan Hanreich wrote:
> The function reading /etc/network/interfaces used /proc/net/dev to
> determine pre-existing physical interfaces. Since the introduction of
> altnames, /proc/net/dev returns insufficient information for
> determining if an interface is alr
Thomas Lamprecht writes:
> On Wed, 23 Jul 2025 13:50:15 +0200, Christian Ebner wrote:
>> Commit 90723828 ("api: backup: add no-cache flag to bypass local
>> datastore cache") introduced the additional flag to request bypassing
>> of the datastore cache by the Proxmox Backup Server.
>>
>> The fla
Am 23.07.25 um 17:16 schrieb Stefan Hanreich:
> On 7/23/25 17:06, Thomas Lamprecht wrote:
>
> [snip]
>
>> I do not recall for sure anymore, but do differing bridge-ports work
>> transparently with the ifupdown2 changes from Christoph. With that it might
>> be
>> nice to support it here too in th
The value returned by get_node_kv is a JSON string and has to be decoded
to read the version of the node.
Signed-off-by: Michael Köppl
---
src/PVE/HA/Env/PVE2.pm | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/src/PVE/HA/Env/PVE2.pm b/src/PVE/HA/Env/PVE2.pm
index aecffc0..
These patches are meant to extend the first part of v3 of the HA rules
series [0]. During testing, I encountered a couple of errors that broke
the migration and also found that there was no UI for the rules. I
reused parts of the v2 series to supply a UI for node affinity rules. I
marked the corres
Signed-off-by: Michael Köppl
---
src/PVE/API2/HA/Rules.pm | 1 +
1 file changed, 1 insertion(+)
diff --git a/src/PVE/API2/HA/Rules.pm b/src/PVE/API2/HA/Rules.pm
index 2e5e382..817aa1d 100644
--- a/src/PVE/API2/HA/Rules.pm
+++ b/src/PVE/API2/HA/Rules.pm
@@ -25,6 +25,7 @@ my $get_api_ha_rule = sub
Add the rules overview that replaces the groups overview and displays
node affinity rules. In addition, the edit dialogs for node affinity
rules are added, allowing both creation and editing of node affinity
rules.
Signed-off-by: Michael Köppl
Originally-by: Daniel Kral
---
This is based on the
Prepend /etc/pve/ to the path, as is done for other read and write
functions for the groups.cfg file. Otherwise, removing the file does not
work.
Signed-off-by: Michael Köppl
---
src/PVE/HA/Config.pm | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/PVE/HA/Config.pm b/src
When exclude_disabled_rules was enabled, rules would be excluded from
the check even if the value was actually false, since it still existed.
The check now makes sure that the rule is really disabled before
excluding it.
Signed-off-by: Michael Köppl
---
src/PVE/HA/Rules.pm | 2 +-
1 file changed
On 7/22/25 18:38, Michael Köppl wrote:
> The migrations based on the persisted rules seemed to work (apart from
> failback). Of course, without the second series, the current state is
> that the groups are migrated and then are simply gone and there is
> nothing to replace them, so it's not usable
Superseded-by:
https://lore.proxmox.com/pve-devel/20250723153524.288508-1-m.koe...@proxmox.com
On 7/22/25 18:54, Michael Köppl wrote:
> These 2 patches are meant as a follow-up for the HA affinity rules
> series [0]. During testing, I encountered some errors during the
> persistent migration that
On Tue, 22 Jul 2025 16:52:21 +0200, Stefan Hanreich wrote:
> Instead of generating a .link file for every virtual function, with this patch
> series proxmox-network-interface-pinning now includes a udev rule that
> dynamically changes the name of virtual functions according to the pinned name
> of
On Tue, 22 Jul 2025 17:14:52 +0200, Stefan Hanreich wrote:
> The key in the ip link output is actually called linkinfo. Before this
> patch, members of bond interfaces that inherit the MAC address of the
> bond would have a wrong MAC in their generated .link file.
>
>
Applied, thanks!
[1/1] net
On 7/23/25 17:06, Thomas Lamprecht wrote:
[snip]
> I do not recall for sure anymore, but do differing bridge-ports work
> transparently with the ifupdown2 changes from Christoph. With that it might be
> nice to support it here too in the midterm, but that is certainly not a
> blocker
> for now.
the following patch seems applicable and might fix an issue observed
in our enterprise support a while ago. containers run in their own
cgroups, thus were probably not scanned by the kernel shrinker - this
resulted in Dnode cache numbers of 300+% reported in arc_summary.
FWICT the issue was introd
On Wed, 23 Jul 2025 16:16:19 +0200, Stoiko Ivanov wrote:
> the directories used to be created in validate_prestart() of the
> sysv-initscript
>
>
Applied, thanks!
[1/1] rrdtool: d/rrdcached.dirs: create /var/lib/rrdcached directories
commit: 2b3eaa79e7b51fc71910c806e25384a5503454cf
it seems to me that the patch does not apply on the latest master. Does
this patch require some other patch be build upon?
On 7/23/25 1:11 PM, Fiona Ebner wrote:
___
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/m
-->8 snip 8<--
>
> -PVE::Tools::file_set_contents($pwfile, "$password\n");
> +PVE::Tools::file_set_contents($pwfile, "$password\n", undef, 1);
i know this is pre-existing, but i'd feel more comfortable forcing the
permissions here rather than depending on the default behaviour. this is
a p
When a cidr address in a FRR access-list is not canonicalized (i.e. is
not a network address) then we get a warning in the journal and
frr-reload.py will even fail. So we need to convert the address entered
by the user into a network address. Factor out the already existing
helper and add a new met
FRR spits out a warning if the address of the access-list is not
canonicalized (i.e. not a network address). So when the user creates a new
fabric, just store the canonicalized version of the passed ip-prefix.
Signed-off-by: Gabriel Goller
---
proxmox-ve-config/src/sdn/fabric/mod.rs | 9 +
On 7/23/25 17:49, Gabriel Goller wrote:
> When frr isn't installed (happens e.g. on pve8 to 9 update) then the SDN
> and Networking "Apply Configuration" task will throw an error because it
> can't access the frr daemons file to check if the daemons are enabled.
> In order to fix this, check if the
On Wed Jul 23, 2025 at 2:17 PM CEST, Michael Köppl wrote:
> The changes good to me. I think this also improves clarity and I did not
> notice any typos or grammatical errors.
>
> Reviewed-by: Michael Köppl
Thanks for the review, Michael.
Would you mind checking out the equivalent patch I sent fo
The function reading /etc/network/interfaces used /proc/net/dev to
determine pre-existing physical interfaces. Since the introduction of
altnames, /proc/net/dev returns insufficient information for
determining if an interface is already contained in /e/n/i, since it
does not include altnames.
The
The lxc-syscalld now uses a different runtime directory. Its old
runtime directory was /run/pve, which was also used for other things,
e.g. storage tunnel and mtunnel sockets and container stderr logs as
well as pull metrics. The fact that it would be recreated on service
restart is problematic, so
When the service is restarted, the directory will be recreated. The
issue is that the /run/pve directory is not just used as the runtime
directory of this service, but also for other things, e.g. storage
tunnel and mtunnel sockets and container stderr logs as well as pull
metrics, which will be los
The pve-lxc-syscalld systemd service currently uses /run/pve as a
runtime directory. This means, that when the service is restarted, the
directory will be recreated. But the /run/pve directory is not just
used as the runtime directory of this service, but also for other
things, e.g. storage tunnel
Changes in v2:
* rebase on current master
* rely on debhelper v13/dh_installtmpfiles
* use root:root 0750 permissions for /run/pve
* avoid redundancy in path for socket
The pve-lxc-syscalld systemd service currently uses /run/pve as a
runtime directory. This means, that when the service is restart
This allows containers started with a configuration using the old
socket path to continue calling in to the pve-lxc-syscalld, except for
a brief time window after the new runtime dir is used before the
postinst script runs. However, such a time window was/is already
present during service restart (
Am 23.07.25 um 3:12 PM schrieb Thomas Lamprecht:
> Am 23.07.25 um 14:56 schrieb Hannes Duerr:
>> it seems to me that the patch does not apply on the latest master. Does this
>> patch require some other patch be build upon?
>
> It's mail. Some lines of the tests are longer than 998 and thus the ma
With the switch to '-blockdev', the 'device' property in the return
value of the 'query-blockstats' QMP command is not initialized. This
can be seen in the QEMU source code (the device property is the name
of the block backend and blk->name is assigned a value only in a code
path reached via drive_
With the switch to '-blockdev', it is necessary to explicitly specify
the device ID for SCSI drives, see also [0]. Otherwise, the device
paths in '/dev/disk/by-id' would be missing. For IDE/SATA, the paths
are still present with '-blockdev' without additional changes and
VirtIO block did not have s
Am 23.07.25 um 3:21 PM schrieb Fiona Ebner:
> With the switch to '-blockdev', it is necessary to explicitly specify
> the device ID for SCSI drives, see also [0]. Otherwise, the device
> paths in '/dev/disk/by-id' would be missing. For IDE/SATA, the paths
> are still present with '-blockdev' withou
On Wed, 23 Jul 2025 13:10:46 +0200, Fiona Ebner wrote:
> With the switch to '-blockdev', it is necessary to explicitly specify
> the device ID for SCSI drives, see also [0]. Otherwise, the device
> paths in '/dev/disk/by-id' would be missing. For IDE/SATA, the paths
> are still present with '-block
When frr isn't installed (happens e.g. on pve8 to 9 update) then the SDN
and Networking "Apply Configuration" task will throw an error because it
can't access the frr daemons file to check if the daemons are enabled.
In order to fix this, check if the frr config file exists before
continuing with t
On 7/23/25 17:33, Thomas Lamprecht wrote:
> Am 23.07.25 um 17:16 schrieb Stefan Hanreich:
>> On 7/23/25 17:06, Thomas Lamprecht wrote:
>>
>> [snip]
>>
>>> I do not recall for sure anymore, but do differing bridge-ports work
>>> transparently with the ifupdown2 changes from Christoph. With that i
Am 23.07.25 um 17:53 schrieb Stefan Hanreich:
> see above, fully supporting the current ifupdown2 capabilities will
> require untangling.
Ack, thanks for looking into this. Might be best done with switching
more of this to rust, ideally with some more tests added upfront and
using the same test as
Am 23.07.25 um 14:56 schrieb Hannes Duerr:
> it seems to me that the patch does not apply on the latest master. Does this
> patch require some other patch be build upon?
It's mail. Some lines of the tests are longer than 998 and thus the mail
server/list hard wraps them, making the diff invalid.
Hi Alexandre,
Am 23.07.25 um 2:15 PM schrieb DERUMIER, Alexandre via pve-devel:
> Hi Fiona, I'm on holiday, can't verify, but before using qemu-img
> measure I had implemented compute of metadatas size (It was not 100%
> perfect).
>
> This is strange, because I thinked that "qemu-img measure" was
Am 23.07.25 um 18:26 schrieb Stefan Hanreich:
>> +if (!-e $PVE::Network::SDN::Frr::FRR_CONFIG_FILE) {
>> +print "frr is not installed, not generating any frr
>> configuration\n";
>> +return;
>> +}
>> +
> The old code checked for the existence of /etc/frr [1] - maybe we shou
The changes good to me. I think this also improves clarity and I did not
notice any typos or grammatical errors.
Reviewed-by: Michael Köppl
On 7/23/25 11:54, Lukas Wagner wrote:
> The latest updates to the backup-job UI completely drop the term
> "Notification System" from the UI, instead we now
--- Begin Message ---
>>Currently, the
>>option does not seem to make a difference with 'qemu-img measure', so
>>that needs to be further investigated.
it seem to be normal that preallocation option don't make a difference
here. qemu-img measure always compute size of all metadatas. (only
cluste
Thanks for the explanation and the hassle with the V2.
And also for @Thomas for appying it, I will go ahead and test it.
On 7/23/25 3:38 PM, Fiona Ebner wrote:
Am 23.07.25 um 3:33 PM schrieb Fiona Ebner:
Am 23.07.25 um 3:21 PM schrieb Fiona Ebner:
With the switch to '-blockdev', it is necessar
the directories used to be created in validate_prestart() of the
sysv-initscript
Signed-off-by: Stoiko Ivanov
---
tested by building this on my bookworm workstation (manually adapting
the build-depends) and checking `dpkg -c ` for the 2 directories
pkgs/rrdtool/rrdtool-1.7.2/debian/rrdcached.dir
On Wed, 23 Jul 2025 17:49:34 +0200, Gabriel Goller wrote:
> When frr isn't installed (happens e.g. on pve8 to 9 update) then the SDN
> and Networking "Apply Configuration" task will throw an error because it
> can't access the frr daemons file to check if the daemons are enabled.
> In order to fix
On Wed, 23 Jul 2025 20:14:53 +0200, Stoiko Ivanov wrote:
> the following patch seems applicable and might fix an issue observed
> in our enterprise support a while ago. containers run in their own
> cgroups, thus were probably not scanned by the kernel shrinker - this
> resulted in Dnode cache numb
On Tue, 22 Jul 2025 14:02:20 +0200, Lukas Wagner wrote:
> Showing warnings for any email address that could not be looked up seems
> appropriate. If the final recipient list is completely empty (as in,
> attempting to send an email with no recipients), an error is still
> raised.
>
>
Applied, th
Moayad Almalat writes:
> ---
> ar.po | 1055 +++--
> 1 file changed, 419 insertions(+), 636 deletions(-)
Thanks, applied at 9154934c251d31772ea4f0731d86520f4606a882.
--
Maximiliano
___
pve-deve
On Mon Jul 21, 2025 at 1:35 PM CEST, Maximiliano Sandoval wrote:
> translator editors and will appear in po files as
>
> ```
> #. TRANSLATORS: Stands for Complete Sequence Number Packet, see
> #.
> https://datatracker.ietf.org/doc/html/draft-ietf-lsr-distoptflood#name-flooding-failures
> #: pve-ma
The latest updates to the backup-job UI completely drop the term
"Notification System" from the UI, instead we now use "Global
notification settings", which should be hopefully a bit clearer to users
with regards to what this actually means.
Furthermore, the 'auto' notification mode is not exposed
On Mon Jul 21, 2025 at 4:44 PM CEST, Maximiliano Sandoval wrote:
> Signed-off-by: Maximiliano Sandoval
> ---
> Differences from v1:
> - s/live guest migration/guest migration/
>
> pvesr.adoc | 18 +-
> 1 file changed, 13 insertions(+), 5 deletions(-)
>
> diff --git a/pvesr.adoc b
This should make it so that the output from `make update` is stable
across different machines.
>From sort(1):
*** WARNING *** The locale specified by the environment affects sort
order. Set LC_ALL=C to get the traditional sort order that uses
native byte values.
LC_COLLATE is a versi
On Tue, Jul 22, 2025 at 04:17:23PM +0200, Max R. Carrara wrote:
> On Tue Jul 22, 2025 at 3:03 PM CEST, Wolfgang Bumiller wrote:
> > applied series thanks
> >
> > (fixed up the doc nits and dropped the make-tidy patch on qemu-server as
> > that was already fixed up in git)
>
> Is there a specific r
I tested this patch series on a fully up-to-date Proxmox VE 8.4.5
cluster of 3 nodes which I then updated to 9.0 BETA. I tested:
- Having both patched and unpatched nodes
- Migrating VMs from patched nodes to unpatched nodes.
- Having only patched nodes.
My test results and remarks:
1. The upd
In commit c26b474 the if condition was removed in favor of the if guard
on the higher level. However, this introduced functional change, the
watchdog now updates immediately after update_watchdog is set to 0,
potentially causing fencing after x+1 seconds instead of x seconds. This
behavior is incor
Maximiliano Sandoval writes:
> In commit c26b474 the if condition was removed in favor of the if guard
> on the higher level. However, this introduced functional change, the
> watchdog now updates immediately after update_watchdog is set to 0,
> potentially causing fencing after x+1 seconds inste
With the switch to '-blockdev', it is necessary to explicitly specify
the device ID for SCSI drives, see also [0]. Otherwise, the device
paths in '/dev/disk/by-id' would be missing. For IDE/SATA, the paths
are still present with '-blockdev' without additional changes and
VirtIO block did not have s
With the switch to '-blockdev', the 'device' property in the return
value of the 'query-blockstats' QMP command is not initialized. This
can be seen in the QEMU source code (the device property is the name
of the block backend and blk->name is assigned a value only in a code
path reached via drive_
--- Begin Message ---
Hi Fiona, I'm on holiday, can't verify, but before using qemu-img
measure I had implemented compute of metadatas size (It was not 100%
perfect).
This is strange, because I thinked that "qemu-img measure" was working
correctly (we need to pass it blocksize && l2_extended optio
Ping
On Mon May 12, 2025 at 5:10 PM CEST, Shan Shaji wrote:
> This patch enables the noVNC console view on iOS devices. It also
> includes a refactor of the function responsible for displaying the
> webview. Additionally, an `AppBar` has been added to the console view on
> iOS to allow users to ea
The 'disk-size' format in the JSON schema in pve-common was changed to
print more verbose, non-ambiguous disk size suffixes. This requires
adapting some tests here.
Bump the corresponding build-dependency in d/control.
Signed-off-by: Fiona Ebner
---
NOTE: check that version bump is correct when
Changes in v3:
* re-add patches for PVE 9
* rebase on current master
The single-letter suffixes are ambiguous and especially in the context
of disks, the powers of ten are usually used. Proxmox VE uses
multiples of 1024 however. By adapting format_size() to use the verbose
suffixes all newly writt
The single-letter suffixes are ambiguous and especially in the context
of disks, the powers of ten are usually used. Proxmox VE uses
multiples of 1024 however. By adapting format_size() to use the verbose
suffixes all newly written or re-written sizes will use the more
verbose/explicit suffix.
Sig
The single-letter suffixes are ambiguous and especially in the context
of disks, the powers of ten are usually used. Proxmox VE uses
multiples of 1024 however.
This is in preparation to adapt format_size() to prefer the verbose
suffixes which is planned for Proxmox VE 9.
The single-letter suffixe
On Wed, 23 Jul 2025 13:50:15 +0200, Christian Ebner wrote:
> Commit 90723828 ("api: backup: add no-cache flag to bypass local
> datastore cache") introduced the additional flag to request bypassing
> of the datastore cache by the Proxmox Backup Server.
>
> The flag is however included in the backu
The 'disk-size' format in the JSON schema in pve-common was changed to
print more verbose, non-ambiguous disk size suffixes. Adapt the tests
to match how configurations are actually written now.
Bump the corresponding build-dependency in d/control.
Signed-off-by: Fiona Ebner
---
NOTE: check tha
On Wed, 23 Jul 2025 10:02:29 +0200, Christian Ebner wrote:
> Two additional patches to improve the documentation for the S3 backend.
>
> The first adds a warning for the costs which might arise from operating
> the S3 backed datastore, the second shows some concrete, provider
> specific configurat
On Wed, 23 Jul 2025 12:25:58 +0200, Christian Ebner wrote:
> This patches improve error logging by refactoring the logging logic
> into a common helper and also covering responses with response status
> BAD_REQUEST. This has not been covered previously and helps to
> diagnose the reason for the bad
The single-letter suffixes are ambiguous and especially in the context
of disks, the powers of ten are usually used. Use exact suffixes
instead when logging the archive file size.
Signed-off-by: Fiona Ebner
---
PVE/VZDump.pm | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --g
69 matches
Mail list logo
