|
2299 | 2299 | If an error is logged then it should be taken seriously as it may require |
2300 | 2300 | user intervention to solve. |
2301 | 2301 |
|
| 2302 | +- name: arangodb_logger_messages_dropped_total |
| 2303 | + introducedIn: "3.12.1" |
| 2304 | + help: | |
| 2305 | + Total number of dropped log messages. |
| 2306 | + unit: number |
| 2307 | + type: counter |
| 2308 | + category: Statistics |
| 2309 | + complexity: simple |
| 2310 | + exposedBy: |
| 2311 | + - agent |
| 2312 | + - coordinator |
| 2313 | + - dbserver |
| 2314 | + - single |
| 2315 | + description: | |
| 2316 | + Total number of log messages that were dropped by the logger. |
| 2317 | + |
| 2318 | + Log messages that are produced by worker threads are pushed into a log |
| 2319 | + queue that is owned by a dedicated logger thread. This dedicated log |
| 2320 | + thread is then responsible for writing the log messages out to disk or |
| 2321 | + stdout asynchronously. |
| 2322 | + As there can be many threads that produce log messages, but only one |
| 2323 | + thread that consumes and logs messages from the log queue, it is possible |
| 2324 | + that the log queue grows too large, potentially consuming a lot of |
| 2325 | + memory to buffer all non-consumed log messages in RAM. |
| 2326 | + In order to prevent the log queue from using too much memory, its size |
| 2327 | + is bounded. The startup option `--log.max-queued-entries` controls the |
| 2328 | + maximum size of the log queue. |
| 2329 | + Whenever the log queue has reached its capacity limit and a log producer |
| 2330 | + thread tries to push a log message to the log queue, the log message |
| 2331 | + will be dropped. For every dropped log message this metric will be |
| 2332 | + increased by one, so it becomes visible that log messages were dropped. |
| 2333 | + troubleshoot: | |
| 2334 | + In case this metric is greater than one, there is too much logging |
| 2335 | + going on for the log thread to keep up with incoming log messages. There |
| 2336 | + are two ways to counteract this. |
| 2337 | + Either the log levels needed to be lowered (e.g. from "trace" to "debug" |
| 2338 | + or from "debug" to "info"), or the maximum capacity of the log queue |
| 2339 | + needs to be increased. This can be achieved by adjusting the value of |
| 2340 | + the startup option `--log.max-queued-entries` to a higher value. |
| 2341 | +
|
2302 | 2342 | - name: arangodb_logger_warnings_total |
2303 | 2343 | introducedIn: "3.9.0" |
2304 | 2344 | help: | |
|
0 commit comments