Change the default log config to reduce disk I/O and storage (#8040)

* Change default log config to buffer by default.

This batches up writes to the filesystem, which is more efficient for
disk I/O. This means that it can take some time for logs to get written
to disk. Note that ERROR logs (and above) immediately flush the buffer.

This only effects new installs, as we only write the log config if
started with `--generate-config` (in the same way we do for generating
signing keys).

* Default to keeping last 4 days of logs.

This hopefully reduces the amount of logs kept for new servers. Keeping
the last 1GB of logs is likely overkill for new servers, but equally may
not be enough for busy ones.

Instead, we keep the last four days worth of logs, enough so that admins
can investigate any problems that happened over e.g. a long weekend.

changelog.d/8040.misc

@@ -0,0 +1 @@
+Change the default log config to reduce disk I/O and storage for new servers.

docs/sample_log_config.yaml

@@ -18,13 +18,29 @@ filters:
 
 handlers:
     file:
-        class: logging.handlers.RotatingFileHandler
+        class: logging.handlers.TimedRotatingFileHandler
         formatter: precise
         filename: /var/log/matrix-synapse/homeserver.log
-        maxBytes: 104857600
+        when: midnight
-        backupCount: 10
+        backupCount: 3  # Does not include the current log file.
-        filters: [context]
         encoding: utf8
+
+    # Default to buffering writes to log file for efficiency. This means that
+    # will be a delay for INFO/DEBUG logs to get written, but WARNING/ERROR
+    # logs will still be flushed immediately.
+    buffer:
+        class: logging.handlers.MemoryHandler
+        filters: [context]
+        target: file
+        # The capacity is the number of log lines that are buffered before
+        # being written to disk. Increasing this will lead to better
+        # performance, at the expensive of it taking longer for log lines to
+        # be written to disk.
+        capacity: 10
+        flushLevel: 30  # Flush for WARNING logs as well
+
+    # A handler that writes logs to stderr. Unused by default, but can be used
+    # instead of "buffer" and "file" in the logger handlers.
     console:
         class: logging.StreamHandler
         formatter: precise
@@ -36,8 +52,23 @@ loggers:
         # information such as access tokens.
         level: INFO
 
+    twisted:
+        # We send the twisted logging directly to the file handler,
+        # to work around https://github.com/matrix-org/synapse/issues/3471
+        # when using "buffer" logger. Use "console" to log to stderr instead.
+        handlers: [file]
+        propagate: false
+
 root:
     level: INFO
-    handlers: [file, console]
+
+    # Write logs to the `buffer` handler, which will buffer them together in memory,
+    # then write them to a file.
+    #
+    # Replace "buffer" with "console" to log to stderr instead. (Note that you'll
+    # also need to update the configuation for the `twisted` logger above, in
+    # this case.)
+    #
+    handlers: [buffer]
 
 disable_existing_loggers: false

synapse/config/logger.py

@@ -62,13 +62,29 @@ filters:
 
 handlers:
     file:
-        class: logging.handlers.RotatingFileHandler
+        class: logging.handlers.TimedRotatingFileHandler
         formatter: precise
         filename: ${log_file}
-        maxBytes: 104857600
+        when: midnight
-        backupCount: 10
+        backupCount: 3  # Does not include the current log file.
-        filters: [context]
         encoding: utf8
+
+    # Default to buffering writes to log file for efficiency. This means that
+    # will be a delay for INFO/DEBUG logs to get written, but WARNING/ERROR
+    # logs will still be flushed immediately.
+    buffer:
+        class: logging.handlers.MemoryHandler
+        filters: [context]
+        target: file
+        # The capacity is the number of log lines that are buffered before
+        # being written to disk. Increasing this will lead to better
+        # performance, at the expensive of it taking longer for log lines to
+        # be written to disk.
+        capacity: 10
+        flushLevel: 30  # Flush for WARNING logs as well
+
+    # A handler that writes logs to stderr. Unused by default, but can be used
+    # instead of "buffer" and "file" in the logger handlers.
     console:
         class: logging.StreamHandler
         formatter: precise
@@ -80,9 +96,24 @@ loggers:
         # information such as access tokens.
         level: INFO
 
+    twisted:
+        # We send the twisted logging directly to the file handler,
+        # to work around https://github.com/matrix-org/synapse/issues/3471
+        # when using "buffer" logger. Use "console" to log to stderr instead.
+        handlers: [file]
+        propagate: false
+
 root:
     level: INFO
-    handlers: [file, console]
+
+    # Write logs to the `buffer` handler, which will buffer them together in memory,
+    # then write them to a file.
+    #
+    # Replace "buffer" with "console" to log to stderr instead. (Note that you'll
+    # also need to update the configuation for the `twisted` logger above, in
+    # this case.)
+    #
+    handlers: [buffer]
 
 disable_existing_loggers: false
 """