I am curious about the separation of the container

[seungjinlee]

I wanted to deeply tune the sentry once, but it seems impossible to delete the various containers in the 20.x.x version in the direction I want. (For example, to completely remove the clickhouse and use only postgres like the previous 9.1.2 version. If I want to do this, I have to modify the python code in the sentry image, right?)
So, I thought about which parts of the onpremise sentry could be tuned. In the end, I came to the conclusion that no tuning was done other than the method of dividing the instance using the environment variable provided by the sentry. (Because I have to fix the inner sentry image to remove the existing container… I really don’t want this)
So, I divided the instances with environment variables, and nginx and postgre were easily separated. But, in the case of Kafka and Snuba, are there any environment variables that provide separate separation? For example, hostname or port. Postgre and nginx have descriptions, so when I set them, they were separated, but it is difficult to find snuba or kafka.
And one more question is, is sentry.config.py used when installing in install.sh? When separating the instance, it seems that some settings should be changed in this sentry.config.py as well as in the container’s environment variable. Is this correct?

[BYK]

@seungjinlee

I wanted to deeply tune the sentry

Can you explain what you mean by “deeply tune”?

@seungjinlee

(For example, to completely remove the clickhouse and use only postgres like the previous 9.1.2 version. If I want to do this, I have to modify the python code in the sentry image, right?)

Yup, you cannot just remove services (containers) and expect things to work normally. We add them for a reason.

@seungjinlee

In the end, I came to the conclusion that no tuning was done other than the method of dividing the instance using the environment variable provided by the sentry. (Because I have to fix the inner sentry image to remove the existing container… I really don’t want this)

Again, I don’t understand what you mean by tuning. We allow customization through the config files under the sentry directory. You can follow https://develop.sentry.dev/config/ for more information on that.

@seungjinlee

So, I divided the instances with environment variables, and nginx and postgre were easily separated. But, in the case of Kafka and Snuba, are there any environment variables that provide separate separation? For example, hostname or port. Postgre and nginx have descriptions, so when I set them, they were separated, but it is difficult to find snuba or kafka.

What do you mean by separation? They are already separate services?

@seungjinlee

And one more question is, is sentry.config.py used when installing in install.sh? When separating the instance, it seems that some settings should be changed in this sentry.config.py as well as in the container’s environment variable. Is this correct?

This file is just some barebones configuration to tell Sentry where the other services are. It indeed is used as otherwise Sentry wouldn’t know how to connect to the database, Snuba, Kafka, Redis, etc. We try to avoid using environment variables, the only ones we pass are

getsentry/self-hosted/blob/19f4561a9e2abe32dc5eb5a03a332b50f2265b4b/docker-compose.yml#L26-L27