From 872609ac98dfe9fce4dd7b97b10c68ada88de555 Mon Sep 17 00:00:00 2001 From: Hemanth Nakkina Date: Fri, 3 Jul 2026 12:23:49 +0530 Subject: [PATCH] docs: add instance-recovery/cluster-refresh snap interface limitation Document LP #2147411 in the known limitations reference, noting that enabling the instance-recovery feature or running a Sunbeam cluster refresh can leave the openstack-network-agents and openstack-hypervisor charms in a blocked state on compute nodes. Add a cross-reference note in the cluster upgrades how-to, scoped to deployments with instance-recovery enabled. Signed-off-by: Hemanth Nakkina --- .custom_wordlist.txt | 1 + how-to/operations/cluster-upgrades.rst | 6 ++++++ reference/known-limitations.rst | 18 +++++++++++++++--- 3 files changed, 22 insertions(+), 3 deletions(-) diff --git a/.custom_wordlist.txt b/.custom_wordlist.txt index 961c934..f9cd2f2 100644 --- a/.custom_wordlist.txt +++ b/.custom_wordlist.txt @@ -182,6 +182,7 @@ idp infrastructure init initiator +instantiation intra ippool isolCPUs diff --git a/how-to/operations/cluster-upgrades.rst b/how-to/operations/cluster-upgrades.rst index e288cdb..c2de896 100644 --- a/how-to/operations/cluster-upgrades.rst +++ b/how-to/operations/cluster-upgrades.rst @@ -112,6 +112,12 @@ charms: sunbeam cluster refresh +.. note:: + If the ``instance-recovery`` feature is enabled, running a cluster refresh + may leave the ``openstack-network-agents`` and ``openstack-hypervisor`` + charms in a blocked state on compute nodes. See + :ref:`Known limitations` for details and the workaround. + If the snap has been refreshed to a different risk level in its channel (for example, from ``stable`` to ``beta``) since the last update, the command will prompt you to confirm before proceeding. In this case, it is recommended diff --git a/reference/known-limitations.rst b/reference/known-limitations.rst index dc7a48c..7d352b7 100644 --- a/reference/known-limitations.rst +++ b/reference/known-limitations.rst @@ -19,6 +19,18 @@ This document describes the known limitations of the Sunbeam project. - Clusters bootstrapped with ``snap-openstack`` versions earlier than **rev998** will continue to experience intermittent API failures (for up to 5 minutes) when a control node becomes unavailable. A known - `k8s charm limitation `_ - restricts updates on active deployments, meaning this issue cannot be - fixed on clusters deployed prior to this revision. \ No newline at end of file + `k8s charm limitation `_ + restricts updates on active deployments, meaning this issue cannot be + fixed on clusters deployed prior to this revision. + * - Enabling the instance-recovery feature blocks the + openstack-network-agents and openstack-hypervisor charms + - `LP #2147411 `_ + - Enabling the ``instance-recovery`` feature can leave the + ``openstack-network-agents`` and ``openstack-hypervisor`` charms in a + blocked state on compute nodes, as the parallel instantiation of + consul-client snaps disrupts the snap interface connections to + MicroOVN ``ovn-chassis`` endpoint. As a workaround, disconnect and + reconnect the ``ovn-chassis`` interface on the + ``openstack-network-agents`` and ``openstack-hypervisor`` snaps, or + reboot the affected host. This issue can also be triggered by running + a Sunbeam cluster refresh.