On restart of a persisted instance, getting "Failed to load instance test-instance from storage: DnaMissing"

[Connoropolous]

Then it goes to
Initializing new chain...

I might be missing something, but I think this is a bug wherein you can't actually reload a whole chain from storage :s

[Connoropolous]

I think the problem is that it doesn't exit when it encounters this kind of error...

holochain-rust/conductor_api/src/conductor/base.rs

Lines 880 to 888 in d832466

	if loading_error == HolochainError::from(NoneError) {
	notify("No chain found in the store".to_string());
	} else {
	notify(format!(
	"Failed to load instance {} from storage: {:?}",
	id.clone(),
	loading_error
	));
	}

[Connoropolous]

@lucksus @AshantiMutinta @zippy can you provide thoughts on this?

[Connoropolous]

In general this makes sense cause I hear that it doesn't actually write the DNA into storage

[jamesray1]

May you please add more repro steps? I agree it looks like unexpected behaviour; on restarting a persisted instance you would expect "Successfully loaded instance test-instance from storage". Maybe could try to debug/check the storage path after Holochain::load(context.clone()).

I found this:

holochain-rust/crates/core/src/persister.rs

Lines 21 to 29 in a5ad16c

	/// trait that defines the persistence functionality that holochain_core requires
	pub trait Persister: Send + Sync {
	// @TODO how does save/load work with snowflake IDs?
	// snowflake is only unique across a single process, not a reboot save/load round trip
	// we'd need real UUIDs for persistant uniqueness
	// @see https://github.com/holochain/holochain-rust/issues/203
	fn save(&mut self, state: &StateWrapper) -> Result<(), HolochainError>;
	fn load(&self, context: Arc<Context>) -> Result<Option<State>, HolochainError>;
	}

Then for further reference:

https://stebalien.github.io/snowflake/snowflake/
"Process unique IDs are guaranteed to be unique within the current process, for the lifetime of the current process."

Link to #203.

Ping @thedavidmeister in case you want to add anything?

[thedavidmeister]

@jamesray1 yes you couldn't safely save any state that is only unique to a given process (e.g. as opposed to UUIDs) - that said i haven't looked around to see if there are still snowflakes in the state (those comments are pretty old)

[jamesray1]

@thedavidmeister LTM like there are still snowflakes, e.g.:

In the network state:

holochain-rust/crates/core/src/network/state.rs

Line 68 in aa53657

id: snowflake::ProcessUniqueId::new(),

In ActionWrapper

https://github.com/holochain/holochain-rust/blob/master/crates/core/src/action.rs#L40-L55

etc:
https://github.com/holochain/holochain-rust/search?q=snowflake&unscoped_q=snowflake

[thedavidmeister]

ok, if these are being saved/loaded it makes sense to me to use UUIDs instead