Merge branch 'rav/fix_expiring_cache_len' into erikj/destination_retry_cache

changelog.d/3911.misc

@@ -0,0 +1 @@
+Fix the docker image building on python 3

changelog.d/3956.bugfix

@@ -0,0 +1 @@
+Fix exceptions from metrics handler

docker/Dockerfile

@@ -1,4 +1,5 @@
-FROM docker.io/python:2-alpine3.8
+ARG PYTHON_VERSION=2
+FROM docker.io/python:${PYTHON_VERSION}-alpine3.8
 
 COPY . /synapse
 

docker/start.py

@@ -5,6 +5,7 @@ import os
 import sys
 import subprocess
 import glob
+import codecs
 
 # Utility functions
 convert = lambda src, dst, environ: open(dst, "w").write(jinja2.Template(open(src).read()).render(**environ))
@@ -23,7 +24,7 @@ def generate_secrets(environ, secrets):
                 with open(filename) as handle: value = handle.read()
             else:
                 print("Generating a random secret for {}".format(name))
-                value = os.urandom(32).encode("hex")
+                value = codecs.encode(os.urandom(32), "hex").decode()
                 with open(filename, "w") as handle: handle.write(value)
             environ[secret] = value
 

synapse/util/caches/__init__.py

@@ -13,6 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import logging
 import os
 
 import six
@@ -20,6 +21,8 @@ from six.moves import intern
 
 from prometheus_client.core import REGISTRY, Gauge, GaugeMetricFamily
 
+logger = logging.getLogger(__name__)
+
 CACHE_SIZE_FACTOR = float(os.environ.get("SYNAPSE_CACHE_FACTOR", 0.5))
 
 
@@ -76,16 +79,20 @@ def register_cache(cache_type, cache_name, cache):
             return []
 
         def collect(self):
-            if cache_type == "response_cache":
-                response_cache_size.labels(cache_name).set(len(cache))
-                response_cache_hits.labels(cache_name).set(self.hits)
-                response_cache_evicted.labels(cache_name).set(self.evicted_size)
-                response_cache_total.labels(cache_name).set(self.hits + self.misses)
-            else:
-                cache_size.labels(cache_name).set(len(cache))
-                cache_hits.labels(cache_name).set(self.hits)
-                cache_evicted.labels(cache_name).set(self.evicted_size)
-                cache_total.labels(cache_name).set(self.hits + self.misses)
+            try:
+                if cache_type == "response_cache":
+                    response_cache_size.labels(cache_name).set(len(cache))
+                    response_cache_hits.labels(cache_name).set(self.hits)
+                    response_cache_evicted.labels(cache_name).set(self.evicted_size)
+                    response_cache_total.labels(cache_name).set(self.hits + self.misses)
+                else:
+                    cache_size.labels(cache_name).set(len(cache))
+                    cache_hits.labels(cache_name).set(self.hits)
+                    cache_evicted.labels(cache_name).set(self.evicted_size)
+                    cache_total.labels(cache_name).set(self.hits + self.misses)
+            except Exception as e:
+                logger.warn("Error calculating metrics for %s: %s", cache_name, e)
+                raise
 
             yield GaugeMetricFamily("__unused", "")