From: Sebastian Wagner <sebastian.wagner@suse.com>
Date: Fri, 20 Mar 2020 14:13:26 +0000 (+0100)
Subject: mgr/cephadm: in-memory set of offline hosts
X-Git-Tag: v15.2.2~76^2~6
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=f6e6aae49eb7e2adaeabe9abcf7639455ff5a14f;p=ceph.git

mgr/cephadm: in-memory set of offline hosts

Fixes: https://tracker.ceph.com/issues/43839
Signed-off-by: Sebastian Wagner <sebastian.wagner@suse.com>
(cherry picked from commit 05cf5796f36bc2d826a29aff4a124124df87cbc6)
---

diff --git a/src/pybind/mgr/cephadm/module.py b/src/pybind/mgr/cephadm/module.py
index 5d3e4c03cdec..d78986a27a90 100644
--- a/src/pybind/mgr/cephadm/module.py
+++ b/src/pybind/mgr/cephadm/module.py
@@ -707,6 +707,9 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
             if h not in self.inventory:
                 self.cache.rm_host(h)
 
+        # in-memory only.
+        self.offline_hosts: Set[str] = set()
+
     def shutdown(self):
         self.log.debug('shutdown')
         self._worker_pool.close()
@@ -1306,6 +1309,11 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
             conn.exit()
         self._cons = {}
 
+    def offline_hosts_remove(self, host):
+        if host in self.offline_hosts:
+            self.offline_hosts.remove(host)
+
+
     @staticmethod
     def can_run():
         if remoto is not None:
@@ -1532,6 +1540,8 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
         if not addr and host in self.inventory:
             addr = self.inventory[host].get('addr', host)
 
+        self.offline_hosts_remove(host)
+
         try:
             try:
                 conn, connr = self._get_connection(addr)
@@ -1616,6 +1626,7 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
             # this is a misleading exception as it seems to be thrown for
             # any sort of connection failure, even those having nothing to
             # do with "host not found" (e.g., ssh key permission denied).
+            self.offline_hosts.add(host)
             user = 'root' if self.mode == 'root' else 'cephadm'
             msg = f'Failed to connect to {host} ({addr}).  ' \
                   f'Check that the host is reachable and accepts connections using the cephadm SSH key\n' \
@@ -1654,6 +1665,7 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
         self.inventory[spec.hostname] = spec.to_json()
         self._save_inventory()
         self.cache.prime_empty_host(spec.hostname)
+        self.offline_hosts_remove(spec.hostname)
         self.event.set()  # refresh stray health check
         self.log.info('Added host %s' % spec.hostname)
         return "Added host '{}'".format(spec.hostname)
@@ -1700,7 +1712,7 @@ class CephadmOrchestrator(orchestrator.Orchestrator, MgrModule):
                 hostname,
                 addr=info.get('addr', hostname),
                 labels=info.get('labels', []),
-                status=info.get('status', ''),
+                status='Offline' if hostname in self.offline_hosts else info.get('status', ''),
             ))
         return r