[ceph_mgr] Update plugin for newer versions of Ceph

Updates the plugin to account for newer versions of Ceph, similar to the
previous few commits focusing on the ceph plugins.

Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>

1 files changed, 73 insertions, 45 deletions

diff --git a/sos/report/plugins/ceph_mgr.py b/sos/report/plugins/ceph_mgr.py
index 2aaf1a9d..a7391f96 100644
--- a/sos/report/plugins/ceph_mgr.py
+++ b/sos/report/plugins/ceph_mgr.py
@@ -6,28 +6,67 @@
 #
 # See the LICENSE file in the source distribution for further information.
 
+import os
+
 from sos.report.plugins import Plugin, RedHatPlugin, UbuntuPlugin
 
 
 class CephMGR(Plugin, RedHatPlugin, UbuntuPlugin):
+    """
+    This plugin is for capturing information from Ceph mgr nodes. While the
+    majority of this plugin should be version-agnostic, several collections are
+    dependent upon the version of Ceph installed. Versions that correlate to
+    RHCS 4 or RHCS 5 are explicitly handled for differences such as those
+    pertaining to log locations on the host filesystem.
+
+    Note that while this plugin will activate based on the presence of Ceph
+    containers, commands are run directly on the host as those containers are
+    often not configured to successfully run the `ceph` commands collected by
+    this plugin. These commands are majorily `ceph daemon` commands that will
+    reference discovered admin sockets under /var/run/ceph.
+    """
 
     short_desc = 'CEPH mgr'
 
     plugin_name = 'ceph_mgr'
     profiles = ('storage', 'virt', 'container')
-    files = ('/var/lib/ceph/mgr/',)
-    containers = ('ceph-(.*)?mgr.*',)
+    files = ('/var/lib/ceph/mgr/', '/var/lib/ceph/*/mgr*')
+    containers = ('ceph-(.*-)?mgr.*',)
 
     def setup(self):
+
+        self.ceph_version = self.get_ceph_version()
+
+        logdir = '/var/log/ceph'
+        libdir = '/var/lib/ceph'
+        rundir = '/run/ceph'
+
+        if self.ceph_version >= 16:
+            logdir += '/*'
+            libdir += '/*'
+            rundir += '/*'
+
         self.add_file_tags({
-            '/var/log/ceph/ceph-mgr.*.log': 'ceph_mgr_log',
+            f'{logdir}/ceph-mgr.*.log': 'ceph_mgr_log',
         })
 
+        self.add_forbidden_path([
+            "/etc/ceph/*keyring*",
+            f"{libdir}/*keyring*",
+            f"{libdir}/**/*keyring*",
+            f"{libdir}/osd*",
+            f"{libdir}/mon*",
+            # Excludes temporary ceph-osd mount location like
+            # /var/lib/ceph/tmp/mnt.XXXX from sos collection.
+            f"{libdir}/tmp/*mnt*",
+            "/etc/ceph/*bindpass*",
+        ])
+
         self.add_copy_spec([
-            "/var/log/ceph/ceph-mgr*.log",
-            "/var/lib/ceph/mgr/",
-            "/var/lib/ceph/bootstrap-mgr/",
-            "/run/ceph/ceph-mgr*",
+            f"{logdir}/ceph-mgr*.log",
+            f"{libdir}/mgr*",
+            f"{libdir}/bootstrap-mgr/",
+            f"{rundir}/ceph-mgr*",
         ])
 
         # more commands to be added later
@@ -42,7 +81,7 @@ class CephMGR(Plugin, RedHatPlugin, UbuntuPlugin):
             "ceph log last cephadm"
         ])
 
-        ceph_cmds = [
+        cmds = [
             "config diff",
             "config show",
             "dump_cache",
@@ -61,44 +100,33 @@ class CephMGR(Plugin, RedHatPlugin, UbuntuPlugin):
             "version"
         ]
 
-        self.add_forbidden_path([
-            "/etc/ceph/*keyring*",
-            "/var/lib/ceph/*keyring*",
-            "/var/lib/ceph/*/*keyring*",
-            "/var/lib/ceph/*/*/*keyring*",
-            "/var/lib/ceph/osd",
-            "/var/lib/ceph/mon",
-            # Excludes temporary ceph-osd mount location like
-            # /var/lib/ceph/tmp/mnt.XXXX from sos collection.
-            "/var/lib/ceph/tmp/*mnt*",
-            "/etc/ceph/*bindpass*",
-        ])
+        self.add_cmd_output([
+            f"ceph daemon {m} {cmd}" for m in self.get_socks() for cmd in cmds]
+        )
 
-        mgr_ids = []
-        # Get the ceph user processes
-        out = self.exec_cmd('ps -u ceph -o args')
-
-        if out['status'] == 0:
-            # Extract the OSD ids from valid output lines
-            for procs in out['output'].splitlines():
-                proc = procs.split()
-                # Locate the '--id' value
-                if proc and proc[0].endswith("ceph-mgr"):
-                    try:
-                        id_index = proc.index("--id")
-                        mgr_ids.append("mgr.%s" % proc[id_index+1])
-                    except (IndexError, ValueError):
-                        self.log_warn("could not find ceph-mgr id: %s", procs)
-
-        # If containerized, run commands in containers
-        try:
-            cname = self.get_all_containers_by_regex("ceph-mgr*")[0][1]
-        except Exception:
-            cname = None
+    def get_ceph_version(self):
+        ver = self.exec_cmd('ceph --version')
+        if ver['status'] == 0:
+            try:
+                _ver = ver['output'].split()[2]
+                return int(_ver.split('.')[0])
+            except Exception as err:
+                self._log_debug(f"Could not determine ceph version: {err}")
+        self._log_error(
+            'Failed to find ceph version, command collection will be limited'
+        )
+        return 0
 
-        self.add_cmd_output([
-            "ceph daemon %s %s"
-            % (mgrid, cmd) for mgrid in mgr_ids for cmd in ceph_cmds
-        ], container=cname)
+    def get_socks(self):
+        """
+        Find any available admin sockets under /var/run/ceph (or subdirs for
+        later versions of Ceph) which can be used for ceph daemon commands
+        """
+        ceph_sockets = []
+        for rdir, dirs, files in os.walk('/var/run/ceph/'):
+            for file in files:
+                if file.startswith('ceph-mgr') and file.endswith('.asok'):
+                    ceph_sockets.append(self.path_join(rdir, file))
+        return ceph_sockets
 
 # vim: set et ts=4 sw=4 :