Skip to main content

Allocation profiling

ClickHouse uses jemalloc as its global allocator that comes with some tools for allocation sampling and profiling.
To make allocation profiling more convenient, SYSTEM commands are provided along 4LW commands in Keeper.

Sampling allocations and flushing heap profiles

If we want to sample and profile allocations in jemalloc, we need to start ClickHouse/Keeper with profiling enabled using environment variable MALLOC_CONF.


jemalloc will sample allocation and store the information internally.

We can tell jemalloc to flush current profile by running:


By default, heap profile file will be generated in /tmp/jemalloc_clickhouse._pid_._seqnum_.heap where _pid_ is the PID of ClickHouse and _seqnum_ is the global sequence number for the current heap profile.
For Keeper, the default file is /tmp/jemalloc_keeper._pid_._seqnum_.heap following the same rules.

A different location can be defined by appending the MALLOC_CONF environment variable with prof_prefix option.
For example, if we want to generate profiles in /data folder where the prefix for filename will be my_current_profile we can run ClickHouse/Keeper with following environment variable:


Generated file will append to prefix PID and sequence number.

Analyzing heap profiles

After we generated heap profiles, we need to analyze them.
For that, we need to use jemalloc's tool called jeprof which can be installed in multiple ways:

  • installing jemalloc using system's package manager
  • cloning jemalloc repo and running from the root folder that will provide you with jeprof script inside the bin folder

jeprof uses addr2line to generate stacktraces which can be really slow.
If that’s the case, we recommend installing an alternative implementation of the tool.

git clone
cd addr2line
cargo b --examples -r
cp ./target/release/examples/addr2line path/to/current/addr2line

There are many different formats to generate from the heap profile using jeprof. We recommend to run jeprof --help to check usage and many different options the tool provides.

In general, jeprof command will look like this:

jeprof path/to/binary path/to/heap/profile --output_format [ > output_file]

If we want to compare which allocations happened between 2 profiles we can set the base argument:

jeprof path/to/binary --base path/to/first/heap/profile path/to/second/heap/profile --output_format [ > output_file]

For example:

  • if we want to generate a text file with each procedure written per line:
jeprof path/to/binary path/to/heap/profile --text > result.txt
  • if we want to generate a PDF file with call-graph:
jeprof path/to/binary path/to/heap/profile --pdf > result.pdf

Generating flame graph

jeprof allows us to generate collapsed stacks for building flame graphs.

We need to use --collapsed argument:

jeprof path/to/binary path/to/heap/profile --collapsed > result.collapsed

After that, we can use many different tools to visualize collapsed stacks.

Most popular would be FlameGraph which contains a script called

cat result.collapsed | /path/to/FlameGraph/ --color=mem --title="Allocation Flame Graph" --width 2400 > result.svg

Another interesting tool is speedscope that allows you to analyze collected stacks in a more interactive way.

Controlling allocation profiler during runtime

If ClickHouse/Keeper were started with enabled profiler, they support additional commands for disabling/enabling allocation profiling during runtime. Using those commands, it's easier to profile only specific intervals.

Disable profiler:


Enable profiler:


It's also possible to control the initial state of the profiler by setting prof_active option which is enabled by default.
For example, if we don't want to sample allocations during startup but only after we enable the profiler, we can start ClickHouse/Keeper with following environment variable:


and enable profiler at a later point.

Additional options for profiler

jemalloc has many different options available related to profiler which can be controlled by modifying MALLOC_CONF environment variable. For example, interval between allocation samples can be controlled with lg_prof_sample.
If you want to dump heap profile every N bytes you can enable it using lg_prof_interval.

We recommend to check jemallocs reference page for such options.

Other resources

ClickHouse/Keeper expose jemalloc related metrics in many different ways.


It's important to be aware that none of these metrics are synchronized with each other and values may drift.

System table asynchronous_metrics

FROM system.asynchronous_metrics
WHERE metric ILIKE '%jemalloc%'
FORMAT Vertical


System table jemalloc_bins

Contains information about memory allocations done via jemalloc allocator in different size classes (bins) aggregated from all arenas.



All jemalloc related metrics from asynchronous_metrics are also exposed using Prometheus endpoint in both ClickHouse and Keeper.


jmst 4LW command in Keeper

Keeper supports jmst 4LW command which returns basic allocator statistics.


echo jmst | nc localhost 9181