Postgresql log file replayer for performance tests
pgreplay [parse options] [replay options] [-d level] [infile]
pgreplay -f [parse options] [-o outfile] [-d level] [infile]
pgreplay -r [replay options] [-d level] [infile]
pgreplay reads a PostgreSQL log file (not a WAL file), extracts the SQL statements and executes them in the same order and relative time against a PostgreSQL database cluster. A final report gives you a useful statistical analysis of your workload and its execution.
In the first form, the log file infile is replayed at the time it is read.
With the -f option, pgreplay will not execute the statements, but write them to a \(oqreplay file\(cq outfile that can be replayed with the third form.
With the -r option, pgreplay will execute the statements in the replay file infile that was created by the second form.
If the execution of statements gets behind schedule, warning messages are issued that indicate that the server cannot handle the load in a timely fashion. The idea is to replay a real-world database workload as exactly as possible.
To create a log file that can be parsed by pgreplay, you need to set the following parameters in postgresql.conf:
log_min_messages=error (or more)
log_min_error_statement=log (or more)
log_connections=on
log_disconnections=on
log_line_prefix='%m|%u|%d|%c|' (if you don't use CSV logging)
log_statement='all'
lc_messages must be set to English (encoding does not matter)
bytea_output=escape (from version 9.0 on, only if you want to replay the log on 8.4 or earlier)
The database cluster against which you replay the SQL statements must be a clone of the database cluster that generated the logs from the time immediately before the logs were generated.
pgreplay is useful for performance tests, particularly in the following situations:
*
You want to compare the performance of your PostgreSQL application on different hardware or different operating systems.
*
You want to upgrade your database and want to make sure that the new database version does not suffer from performance regressions that affect you.
Moreover, pgreplay can give you some feeling as to how your application might scale by allowing you to try to replay the workload at a higher speed. Be warned, though, that 500 users working at double speed is not really the same as 1000 users working at normal speed.
-c
Specifies that the log file is in 'csvlog' format (highly recommended) and not in 'stderr' format.
-b timestamp
Only log entries greater or equal to that timestamp will be parsed. The format is YYYY-MM-DD HH:MM:SS.FFF like in the log file. An optional time zone part will be ignored.
-e timestamp
Only log entries less or equal to that timestamp will be parsed. The format is YYYY-MM-DD HH:MM:SS.FFF like in the log file. An optional time zone part will be ignored.
-q
Specifies that a backslash in a simple string literal will escape the following single quote. This depends on configuration options like standard_conforming_strings and is the default for server version 9.0 and less.
-h hostname
Host name where the target database cluster is running (or directory where the UNIX socket can be found). Defaults to local connections.
This works just like the -h option of psql.
-p port
TCP port where the target database cluster can be reached.
-W password
By default, pgreplay assumes that the target database cluster is configured for trust authentication. With the -W option you can specify a password that will be used for all users in the cluster.
-s factor
Speed factor for replay, by default 1. This can be any valid positive floating point number. A factor less than 1 will replay the workload in \(oqslow motion\(cq, while a factor greater than 1 means \(oqfast forward\(cq.
-E encoding
Specifies the encoding of the log file, which will be used as client encoding during replay. If it is omitted, your default client encoding will be used.
-j
If all connections are idle, jump ahead to the next request instead of sleeping. This will speed up replay. Execution delays will still be reported correctly, but replay statistics will not contain the idle time.
-o outfile
specifies the replay file where the statements will be written for later replay.
-d level
Specifies the trace level (between 1 and 3). Increasing levels will produce more detailed information about what pgreplay is doing.
-v
Prints the program version and exits.
PGHOST
Specifies the default value for the -h option.
PGPORT
Specifies the default value for the -p option.
PGCLIENTENCODING
Specifies the default value for the -E option.
pgreplay can only replay what is logged by PostgreSQL. This leads to some limitations:
*
COPY statements will not be replayed, because the copy data are not logged.
*
Fast-path API function calls are not logged and will not be replayed. Unfortunately, this includes the Large Object API.
*
Since the log file is always in the server encoding (which you can specify with the -E switch of pgreplay), all SET client_encoding statements will be ignored.
*
Since the preparation time of prepared statements is not logged (unless log_min_messages is debug2 or more), these statements will be prepared immediately before they are first executed during replay.
*
Because the log file contains only text, query parameters and return values will always be in text and never in binary format. If you use binary mode to, say, transfer large binary data, pgreplay can cause significantly more network traffic than the original run.
*
Sometimes, if a connection takes longer to complete, the session ID unexpectedly changes in the PostgreSQL log file. This causes pgreplay to treat the session as two different ones, resulting in an additional connection. This is arguably a bug in PostgreSQL.
Written by Laurenz Albe <[email protected]>.