Odoo Profiling With py-spy

Odoo already have a tool to profiling the code inside the module. But sometimes we don’t know where the performance issue came from so we can’t just add decorator @profile to every methods we have. In this case, there is a tool called py-spy that can attach to Odoo main process directly.

Installation

For most OSes, we can easily install it using pip

python3 -m pip install py-spy

please follow this installation section from py-spy

Usage

We will restrict py-spy in here to record by attaching to Odoo main process ID and check the flamegraph using Speedscope.app. As for details in usage, you can always check it here

After you have installed py-spy, you can always confirm if it installed by using --help flag

$ py-spy --help
py-spy 0.3.14
Sampling profiler for Python programs

USAGE:
    py-spy <SUBCOMMAND>

OPTIONS:
    -h, --help       Print help information
    -V, --version    Print version information

SUBCOMMANDS:
    record    Records stack trace information to a flamegraph, speedscope or raw file
    top       Displays a top like view of functions consuming CPU
    dump      Dumps stack traces for a target program to stdout
    help      Print this message or the help of the given subcommand(s)
$

Run odoo, in this case, we will run odoo-bin manually inside virtual environment
Open new terminal in new tab (or window), then find the process ID. We can use top, or htop to find the process and write-down the process ID to run odoo (here we use htop and using Search function (F3) to find odoo-bin, and found the process ID is 18203)
Then close htop, and run py-spy record while attach to process ID above (you can always check available flags with py-spy record --help)
```
sudo py-spy record -p 18203 -f speedscope -o odoo-profiling.txt
```
Need to use sudo since the process is attaching to the process ID of odoo
It will show like this when it is recording. While at it, try some function in odoo (like create Purchase Order, Sales Order, or anything you want to check the traces)
```
$ sudo py-spy record -p 18203 -f speedscope -o odoo-profiling.txt
py-spy> Sampling process 100 times a second. Press Control-C to exit.
```

After it’s done, just cancel py-spy process with CTRL+C, and new file will be created at the same directory you run py-spy

$ sudo py-spy record -p 18203 -f speedscope -o odoo-profiling.txt
py-spy> Sampling process 100 times a second. Press Control-C to exit.

^C
py-spy> Stopped sampling because Control-C pressed
py-spy> Wrote speedscope file to 'odoo-profiling.txt'. Samples: 480 Errors: 1
py-spy> Visit https://www.speedscope.app/ to view

Now, open Speedscope.app and open the file created by py-spy. There will be so many threads that you can see in the top-center of the page, check one-by-one and look for the method that you know will run when doing the task. Internal odoo is already big even when you don’t have any module installed. So, it will spawn some thread and run some process in different thread or child-process, take your time to find the method and function you sure will run whenever you did the action.
Also in the image above, we can see a flamegraph. You can check this article from webperf.tips on how to read flamegraph.
For example, we know that everytime we create new record in Purchase Order, then add new Order Line and when choosing a product, method onchange_product_id will run. We can click the process, and it shown new view at the bottom of the page, this new view shown process before this method onchange_product_id called, and if we see the flamegraph again, there is some methods below onchange_product_id, this methods (like _product_id_change) will be called while onchange_product_id running.

we can combine py-spy with internal odoo profiler. Py-spy can help to pinpoint the method that takes too much time, then after we know the process, we can use odoo profiler to check the time spent of the process in every line.

Installation

Usage

Comments