Logging
Introducing Couchbase Sync Gateway’s logging functionality
Sync Gateway’s _Continuous Logging_ feature delivers flexible log generation and retention, without compromising the availability of diagnostic information necessary to provide effective support and maintenance.
Constraints
Do not use the |
Overview
Sync Gateway provides a robust Continuous Logging[1] solution that delivers flexibility in terms of how logs are generated and retained, whilst also maintaining the level of logging required by Couchbase Support for investigation of issues. Its logs are written to separate log files filtered by log level, with each log level supporting individual retention policies. You control what is logged using the Bootstrap Configuration logging properties [fnleg3-0].
In addition to the log files, you can also independently configure and control Console Logging, which is a convenient method of accessing diagnostic information during debugging scenarios. With console logging, system administrators can easily fine-tune diagnostic output to suit specific debug scenarios. All without interfering with the logging required by Couchbase Support for the investigation of issues.
Configuration
You configure continuous and console logging requirements in the Bootstrap Configuration file, using the logging properties [2] — see: Example 1.
{
"logging": {
"log_file_path": "/var/tmp/sglogs", (1)
"redaction_level": "partial", (2)
"console": { (3)
"log_level": "debug",
"log_keys": ["*"]
}, (4)
"error": { (5)
"enabled": true,
"rotation": {
"max_size": 20,
"max_age": 180
}
},
"warn": { (6)
"enabled": true,
"rotation": {
"max_size": 20,
"max_age": 90
}
},
"info": { (7)
"enabled": false
},
"debug": { (8)
"enabled": false
}
}
}
1 | Set the path to the log file(s) |
2 | Define the optional redaction level, here we select "partial" redaction — see: Log Redaction |
3 | Here we define the Console Logging levels we require for debugging. In this instance turning on debug level output for all available log_keys — see: logging.console.log_keys [2] |
4 | The following logging properties enable, or disable, Continuous Logging at the specified level — see: logging.console.log_level [2] |
5 | Here we set error level logging on and define the log-file rotation, for errors messages |
6 | Here we set warn level logging on and define the log-file rotation, for warnings messages |
7 | Here we set info level logging off; we do not define the log-file rotation |
8 | Here we set debug level logging off; we do not define the log-file rotation |
See: Bootstrap Configuration for more information on these settings [2].
Log Redaction
All log outputs — console or continuous — can optionally be redacted, which will remove any user-data considered private.
You enable this feature by setting the logging.redaction_level property [2].
Console Logging
In this section: Log Levels | Admin REST API | Log Keys | Set Log Color | Redirect Console Log
By default only HTTP logging is enabled |
Console logs are your go-to resource for diagnostic information. You can easily fine-tune their diagnostic content to meet the needs of a particular debugging scenario, perhaps by increasing the verbosity and filtering out unnecessary log_keys to better focus on the problem area.
Changes to console logging are independent of continuous logging, so you can, for example, tweak any of the following without compromising the core continuous logging streams:
-
Increase the verbosity using Log Levels to generate additional diagnostic information
-
Focus on the area under investigation by enabling or disabling specific Log Keys
-
Enhance readability by setting a color for log output based on log level
Admin REST API
You can define console log settings in the bootstrap configuration file, or more conveniently, you can use the Admin REST API’s _config
endpoint to set them — see: Example 2.
curl --location --request PUT 'http://localhost:4985/_config' \
--header 'Content-Type: application/json' \
--data-raw '{
"console": {
"log_level": "trace", (1)
"log_keys": "[\"WS\",\"WSFrame\",\"Replicate\"]", (2)
}
}'
1 | Here we define the log_level to be trace for maximum verbosity |
2 | Here we specify the particular log_keys we want to focus on. Note here we have chosen to exclude http ; perhaps we want to focus on a websocket or replication issue. |
The console log will show the following after this command:
2021-01-08T13:26:23.884Z [INF] HTTP: #110: POST /_logging?logLevel=trace (as ADMIN)
2021-01-08T13:26:23.885Z [INF] Setting log level to: trace
2021-01-08T13:26:23.885Z [INF] Setting log keys to: [DCP Replicate WS WSFrame]
Log Levels
When debugging, setting the console log’s log-level to debug or trace can provide valuable additional information
|
Console logs have six levels of verbosity — see: Table 1.
The default log level is none
Note that the log levels are inclusive, so if you enable info
level, then warn
and error
logs are also enabled.
You can define console log levels in the configuration file — see logging.console.log_level [2] and by using the Admin REST API (see: Example 2).
One approach might be to set your base level in the configuration file and then use the Admin REST API for specific debugging scenarios.
Log Level |
Appearance |
Description |
---|---|---|
|
- |
Disables log output |
|
|
Displays errors that need urgent attention |
|
|
Displays warnings that need some attention |
|
|
Displays information about normal operations that don’t need attention |
|
|
Displays verbose output that might be useful when debugging |
|
|
Displays extremely verbose output that might be useful when debugging |
Log Keys
Select log keys relevant to the area you are debugging, providing them as a comma-delimited list, such as: "log_keys": ["HTTP", "CRUD", "Import"] in the config or see Example 2 for how to provide them using the Admin REST API.
|
Log keys provide fine-grained control over the information types that Sync Gateway outputs to the console log.
By default, only HTTP
related information is enabled, but a range of other keys are available to meet specific diagnostic needs — see: Table 2.
You can define the required logging.console.log_keys within your configuration file and-or use the Admin REST API (see: Example 2).
Log Key |
Description |
---|---|
* |
This wildcard log key, enables all log keys
|
none |
Disable all log keys; no logging output
|
Admin |
Admin processes in Sync Gateway. |
Access |
Anytime an access() call is made in the sync function. |
Auth |
Authentication. |
Bucket |
Sync Gateway interactions with the bucket (trace level only). |
Cache |
Interactions with Sync Gateway’s in-memory channel cache. |
Changes |
Processing of /{db}/_changes requests. |
CRUD |
Updates made by Sync Gateway to documents. |
DCP |
DCP-feed processing. |
Events |
Event processing (webhooks). |
gocb |
All logging emitted by the GoCB SDK |
HTTP |
All requests made to the Sync Gateway REST APIs. |
HTTP+ |
Additional information about HTTP requests (response times, status codes). |
Import |
This log key can be useful to troubleshoot why a given document was not successfully imported. |
Javascript |
All logging from Javascript. This includes: sync function, import filters, webhook filter function, and custom inter-Sync Gateway replication conflict resolvers |
Migrate |
Logs messages that show when old inline document metadata is upgraded to xattrs |
Query |
Query is used for Sync Gateway code related to N1QL queries |
Replicate |
Log messages related to replications between Sync Gateways (using sg-replicate). This tag cannot be used for replications initiated by Couchbase Lite. |
SGCluster |
Log messages related to the sharded import and HA sg-replicate |
Sync |
Activity which relates to synchronization between Couchbase Lite and Sync Gateway |
SyncMsg |
Can be used for additional Sync logging output |
WS |
Websocket replication log messages |
WSFrame |
Can be used for additional WS logging output |
Set Log Color
To use color for log output based on log level, set logging.console.color_enabled to true
This setting is always disabled on Windows for compatibility reasons. |
Redirect Console Log
You can easily redirect the console log output to a file. This can be useful not only for diagnostic sessions, but also when you have specialized logging requirements, such as centralized logging. Just redirect the output and then apply your own log collection mechanism to feed that data elsewhere — see Example 3.
# Start Sync Gateway and redirect console output to a file
./sync-gateway > my_sg_logs.txt 2>&1
# Start log collection to send to a centralized log aggregator.
logcollector my_sg_logs.txt
Continuous Logging
In this section: Log File Outputs | Log File Rotation
Continuous logging produces a set of log files aimed primarily at providing appropriate diagnostic information for the Couchbase Support team should their intervention be required. You define continuous logging settings in the configuration file — see: Example 1.
With continuous logging the logs for each level are written to separate log files — see: Table 3. You can set individual retention policies for each log-level.
Log File Outputs
The log files output from continuous logging are intended solely for the use of Couchbase Support.
If you require special log handling, for example for centralized logging, then use the Redirect Console Log feature to create a log file for this purpose from the console output stream. |
Sync Gateway produces four separate log files, split by log level. Each log file has its own guaranteed retention period - as shown in Table 3
You can collect the log files, for analysis by Couchbase Support when diagnosing Sync Gateway issues, using SGCollect Info.
Log File |
Level |
Description |
Default enabled |
Default max_age |
Minimum max_age |
---|---|---|---|---|---|
ERROR |
Critical error messages. |
|
|
360 Days |
180 Days |
WARN |
Something is wrong but SG can still service requests |
|
|
180 Days |
90 Days |
INFO |
Important diagnostics for support and customers |
|
|
6 Days |
3 Days |
DEBUG |
Lower level development analysis |
|
|
2 Days |
1 Day |
Each log level and its parameters are defined using the logging.console.log_level property. |
Log File Rotation
Log files are rotated when they exceed a threshold max_size
(megabytes).
Once rotated, they are compressed (gzip) to reduce the disk usage.
Aged logs are cleaned up once their age exceeds max_age
days — see: Table 3
Configure log rotation using the logging.console.rotation property.
For pre-2.1 log rotation — see: Pre 2.1 Log Rotation