= Request Tracing

Gerrit supports on-demand tracing of single requests that results in
additional logs with debug information that are written to the
`error_log`. The logs that correspond to a traced request are
associated with a unique trace ID. This trace ID is returned with the
response and can be used by an administrator to find the matching log
entries.

How tracing is enabled and how the trace ID is returned depends on the
request type:

* REST API: For REST calls tracing can be enabled by setting the
  `trace` request parameter, the trace ID is returned as
  `X-Gerrit-Trace` header. More information about this can be found in
  the link:rest-api.html#tracing[Request Tracing] section of the
  link:rest-api.html[REST API documentation].
* SSH API: For SSH calls tracing can be enabled by setting the
  `--trace` option. More information about this can be found in
  the link:cmd-index.html#trace[Trace] section of the
  link:cmd-index.html[SSH command documentation].
* Git: For Git pushes tracing can be enabled by setting the
  `trace` push option, the trace ID is returned in the command output.
  More information about this can be found in
  the link:user-upload.html#trace[Trace] section of the
  link:user-upload.html[upload documentation]. Tracing for Git requests
  other than Git push is not supported.

When request tracing is enabled it is possible to provide an ID that
should be used as trace ID. If a trace ID is not provided a trace ID is
automatically generated. The trace ID must be provided to the support
team so that they can find the trace.

When doing traces it is recommended to specify the ID of the issue
that is being investigated as trace ID so that the traces of the issue
can be found more easily. When the issue ID is used as trace ID there
is no need to find the generated trace ID and report it in the issue.

Since tracing consumes additional server resources tracing should only
be enabled for single requests if there is a concrete need for
debugging. In particular bots should never enable tracing for all their
requests by default.

== Find log entries for a trace ID

If tracing is enabled all log messages that correspond to the traced
request have a `TRACE_ID` tag set, e.g.:

----
[2018-08-13 15:28:08,913] [HTTP-76] TRACE com.google.gerrit.httpd.restapi.RestApiServlet : Received REST request: GET /a/accounts/self (parameters: [trace]) [CONTEXT forced=true TRACE_ID="1534166888910-3985dfba" ]
[2018-08-13 15:28:08,914] [HTTP-76] TRACE com.google.gerrit.httpd.restapi.RestApiServlet : Calling user: admin [CONTEXT forced=true TRACE_ID="1534166888910-3985dfba" ]
[2018-08-13 15:28:08,942] [HTTP-76] TRACE com.google.gerrit.httpd.restapi.RestApiServlet : REST call succeeded: 200 [CONTEXT forced=true TRACE_ID="1534166888910-3985dfba" ]
----

By doing a grep with the trace ID over the error log the log entries
that correspond to the request can be found.

== Which information is captured in a trace?

* request details
** REST API: request URL, request parameter names, calling user,
   response code, response body on errors
** SSH API: parameter names
** Git API: push options, magic branch parameter names
* cache misses, cache evictions
* reads from NoteDb, writes to NoteDb
* reads of meta data files, writes of meta data files
* index queries (with parameters and matches)
* reindex events
* permission checks (e.g. which rule is responsible for a deny)
* timer metrics
* all other logs