mgr/cephadm: add info to 'ceph orch upgrade status' in cephadm

properly fills in 'services_complete' field, adds info messages to 'message' field
such as what daemon type is being upgraded or if we're pulling an image and adds
'progress' field that shows how many daemons have been upgraded so far.

Fixes: https://tracker.ceph.com/issues/49235
Signed-off-by: Adam King <adking@redhat.com>
(cherry picked from commit 1527b3a527c45deb689f3d7b5aff02e517357584)

diff --git a/src/pybind/mgr/cephadm/upgrade.py b/src/pybind/mgr/cephadm/upgrade.py
index 24b36fa8fc393c5b230806a0d35c9d87fde4c2ac..fddabac28672763eefe1b58f4203c86afd25370c 100644 (file)
--- a/src/pybind/mgr/cephadm/upgrade.py
+++ b/src/pybind/mgr/cephadm/upgrade.py
@@ -2,7 +2,7 @@ import json
 import logging
 import time
 import uuid
-from typing import TYPE_CHECKING, Optional, Dict, List
+from typing import TYPE_CHECKING, Optional, Dict, List, Tuple
 
 import orchestrator
 from cephadm.serve import CephadmServe
@@ -92,12 +92,38 @@ class CephadmUpgrade:
         if self.upgrade_state:
             r.target_image = self.target_image
             r.in_progress = True
+            r.progress, r.services_complete = self._get_upgrade_info()
+            # accessing self.upgrade_info_str will throw an exception if it
+            # has not been set in _do_upgrade yet
+            try:
+                r.message = self.upgrade_info_str
+            except AttributeError:
+                pass
             if self.upgrade_state.error:
                 r.message = 'Error: ' + self.upgrade_state.error
             elif self.upgrade_state.paused:
                 r.message = 'Upgrade paused'
         return r
 
+    def _get_upgrade_info(self) -> Tuple[str, List[str]]:
+        if not self.upgrade_state or not self.upgrade_state.target_digests:
+            return '', []
+
+        daemons = [d for d in self.mgr.cache.get_daemons() if d.daemon_type in CEPH_UPGRADE_ORDER]
+
+        if any(not d.container_image_digests for d in daemons if d.daemon_type == 'mgr'):
+            return '', []
+
+        completed_daemons = [(d.daemon_type, any(d in self.upgrade_state.target_digests for d in (
+            d.container_image_digests or []))) for d in daemons if d.daemon_type]
+
+        done = len([True for completion in completed_daemons if completion[1]])
+
+        completed_types = list(set([completion[0] for completion in completed_daemons if all(
+            c[1] for c in completed_daemons if c[0] == completion[0])]))
+
+        return '%s/%s ceph daemons upgraded' % (done, len(daemons)), completed_types
+
     def _check_target_version(self, version: str) -> Optional[str]:
         try:
             (major, minor, patch) = version.split('.')
@@ -370,6 +396,7 @@ class CephadmUpgrade:
         if not target_id or not target_version or not target_digests:
             # need to learn the container hash
             logger.info('Upgrade: First pull of %s' % target_image)
+            self.upgrade_info_str = 'Doing first pull of %s image' % (target_image)
             try:
                 target_id, target_version, target_digests = CephadmServe(self.mgr)._get_container_image_info(
                     target_image)
@@ -418,7 +445,7 @@ class CephadmUpgrade:
 
         image_settings = self.get_distinct_container_image_settings()
 
-        daemons = self.mgr.cache.get_daemons()
+        daemons = [d for d in self.mgr.cache.get_daemons() if d.daemon_type in CEPH_UPGRADE_ORDER]
         done = 0
         for daemon_type in CEPH_UPGRADE_ORDER:
             logger.info('Upgrade: Checking %s daemons' % daemon_type)
@@ -456,6 +483,9 @@ class CephadmUpgrade:
             ):
                 return
 
+            if need_upgrade:
+                self.upgrade_info_str = 'Currently upgrading %s daemons' % (daemon_type)
+
             to_upgrade = []
             known_ok_to_stop: List[str] = []
             for d in need_upgrade:
@@ -502,6 +532,8 @@ class CephadmUpgrade:
                 if code or not any(d in target_digests for d in json.loads(''.join(out)).get('repo_digests', [])):
                     logger.info('Upgrade: Pulling %s on %s' % (target_image,
                                                                d.hostname))
+                    self.upgrade_info_str = 'Pulling %s image on host %s' % (
+                        target_image, d.hostname)
                     out, errs, code = CephadmServe(self.mgr)._run_cephadm(
                         d.hostname, '', 'pull', [],
                         image=target_image, no_fsid=True, error_ok=True)
@@ -519,10 +551,14 @@ class CephadmUpgrade:
                     if not any(d in target_digests for d in r.get('repo_digests', [])):
                         logger.info('Upgrade: image %s pull on %s got new digests %s (not %s), restarting' % (
                             target_image, d.hostname, r['repo_digests'], target_digests))
+                        self.upgrade_info_str = 'Image %s pull on %s got new digests %s (not %s), restarting' % (
+                            target_image, d.hostname, r['repo_digests'], target_digests)
                         self.upgrade_state.target_digests = r['repo_digests']
                         self._save_upgrade_state()
                         return
 
+                    self.upgrade_info_str = 'Currently upgrading %s daemons' % (daemon_type)
+
                 if len(to_upgrade) > 1:
                     logger.info('Upgrade: Updating %s.%s (%d/%d)' %
                                 (d.daemon_type, d.daemon_id, num, len(to_upgrade)))

diff --git a/src/pybind/mgr/orchestrator/_interface.py b/src/pybind/mgr/orchestrator/_interface.py
index 29b8603818ea3c4a70a7d1119569866a911ce9c4..11c70b0f8354f32c9cb366e9cb1da09cbeb8849e 100644 (file)
--- a/src/pybind/mgr/orchestrator/_interface.py
+++ b/src/pybind/mgr/orchestrator/_interface.py
@@ -734,6 +734,7 @@ class UpgradeStatusSpec(object):
         self.in_progress = False  # Is an upgrade underway?
         self.target_image: Optional[str] = None
         self.services_complete: List[str] = []  # Which daemon types are fully updated?
+        self.progress: Optional[str] = None  # How many of the daemons have we upgraded
         self.message = ""  # Freeform description
 
 

diff --git a/src/pybind/mgr/orchestrator/module.py b/src/pybind/mgr/orchestrator/module.py
index d314dc27ae11832105f8a418641535b524dab03b..bcffdc833e6c03e93b2e91265e010656aa31dfa8 100644 (file)
--- a/src/pybind/mgr/orchestrator/module.py
+++ b/src/pybind/mgr/orchestrator/module.py
@@ -1298,6 +1298,7 @@ Usage:
             'target_image': status.target_image,
             'in_progress': status.in_progress,
             'services_complete': status.services_complete,
+            'progress': status.progress,
             'message': status.message,
         }
         out = json.dumps(r, indent=4)