Always restart container chain before starting collation #307
Conversation
The user can specify p2p, rpc, and prometheus ports using the CLI, and since there can be 2 container chains running at the same time, the port used may be the one specified by the user, or that +1
tmpolaczyk
force-pushed
the
tomasz-ports-fix-2
branch
from
c4554de to
d15de55
Compare
|
Master coverage: 71.35% |
|
|
| @@ -89,14 +89,16 @@ pub struct ContainerChainSpawnerState { | |||
|
|
|||
| pub struct ContainerChainState { | |||
| /// Async callback that enables collation on this container chain | |||
| // We don't use it since we are always restarting container chains | |||
| #[allow(unused)] | |||
There was a problem hiding this comment.
Interesting should we remove it already then?
There was a problem hiding this comment.
We could, but it will make it a bit harder to revert this change in the future
| if !start_collation { | ||
| log::info!("This is a syncing container chain, using random ports"); | ||
| // Use random ports to avoid conflicts with the other running container chain | ||
| let random_ports = [23456, 23457, 23458]; |
There was a problem hiding this comment.
wont it automatically use random ports even if we dont assign any?, in other words, can we use None here?
There was a problem hiding this comment.
We maybe can use [0, 0, 0], but the reason I selected the random ports myself is that if we actually use random ports there is a small probability of occupying a port from the collating container. The way it is implemented now that will only happen if this fixed random port is already taken, which can happen in zombienet because all of the collators try to use the same ports. But in production where every collator runs in isolation, I think it is better to use these fixed random ports instead of actually random.
| } | ||
|
|
||
| if need_to_restart_next { | ||
| // Handle edge case of going from (2000, 2001) to (2001, 2000). In that case we must restart both chains, |
There was a problem hiding this comment.
in the second case, it restarts with random ports?
There was a problem hiding this comment.
Yes, we stop both of them, wait 10 seconds, and then start the new "syncing" with random ports, and the new "collating" with the good ports (in either order).
* Increment ports of "odd" parachains to avoid port conflicts The user can specify p2p, rpc, and prometheus ports using the CLI, and since there can be 2 container chains running at the same time, the port used may be the one specified by the user, or that +1 * Always restart container chain before starting collation * Fix wrong condition and add keep_db to stop signal * Fix unit tests * Fix edge case where we need to restart the next chain
Container chains can start in "sync mode", which uses random ports, or in "collation mode" which uses the ports provided by the user through CLI. To ensure that the container chain that is currently collating uses the correct ports, we restart it right when it is time to start collation.