Long running Running pocket-ic/dfx replica

Developers Language Support
1

I’m curious about the ability for pocket ic and/or the dfx replica to be a long running process. Say something like trying to really simulate a large subnet of a long period.

I’d pocket-ic in continuous mode basically the replica with without all the consensus turned on, or is it some kind of fundamentally different process?

I’ve been thinking about trying to simulate and process some more robust things with motoko and I’m wondering if I can basically just run pocket-ic as kind of motoko server and I’m curious what kind of sla I’d get, what kind of memory/cpu would be needed?

Would I be better off trying to look at running a replica configured for one node? Is that even possible and/or is that binary available to build? If so, what would be the build process and what would it need to run reliably?

1 Like
2

It is ‘multiple replicas’ (one for each subnet) and not only without consensus, but also a custom (more lightweight) messaging layer, and no p2p layer AFAIK.

No, I don’t think you’d be better off. dfx start basically does basically that, but I have heard quite a few complaints about memory leaks when running the replica for a long time. Also, we’ll remove the replica from dfx in the near future.

3

We haven’t really tested PocketIC in this mode, so I can’t promise any SLA. In particular, it might also be the case that you get the same behavior in terms of memory leaks that Severin described above.