Isolation dir

Every buck2 daemon has an "isolation directory" which acts as an isolation mechanism. This enables the ability to have multiple independent daemons, each with a different isolation dir.

Different daemons enable the concurrent execution of commands, as a single daemon can generally execute only a single command at a time. Most buck2 commands only act in their own specified isolation dir. For example, buck2 kill will only kill the daemon associated to the default isolation-dir (v2). Exceptions to this rule include buck2 killall command, which will kill all buck2 processes on the machine.

WARNING: Isolation dirs have a very important consequence: buck2 invocations with different isolation dirs never share any cached artifacts.

Isolation dir uses

Isolation dirs are also relied on for developer environment tooling (e.g. LSPs) that have to run in the background without affecting the users manually triggered builds.

Also isolation dirs are used to allow for recursive invocations.

How to set the isolation dir for a command

The default isolation dir is v2. Thus any buck2 command without an explicit isolation dir is equivalent to

$ buck2 --isolation-dir v2 $ARGS

The isolation dir can also be set via the BUCK_ISOLATION_DIR env var.

NOTE: --isolation-dir arg must always appear immediately after buck2. For example, buck2 build --isolation-dir v2 $ARGS is an invalid command.

Isolation dir buck-out path

The isolation dir specifies the root directory in buck-out which will be used to isolate a daemon. With the default v2 isolation dir, buck2 will be using the $PROJECT_ROOT/buck-out/v2 directory for all target outputs and internal metadata. $PROJECT_ROOT refers to the directory where the project lives.