Drop Pending Stakers 3 - persist stakers' StartTime #2306
Conversation
abi87
changed the title
abi87
changed the title
abi87
changed the title
| } | ||
|
|
||
| func writeDelegatorMetadata(db database.KeyValueWriter, metadata *delegatorMetadata) error { | ||
| return database.PutUInt64(db, metadata.txID[:], metadata.PotentialReward) | ||
| metadataBytes, err := metadataCodec.Marshal(v1, metadata) |
There was a problem hiding this comment.
we store start time as soon as we deploy code, without waiting for fork activation
There was a problem hiding this comment.
Can people downgrade after updating to a version with this PR?
There was a problem hiding this comment.
The versioning should allow smooth handling of upgrades/downgrades
When reading from disk:
- If metadata have version zero, start time won't be loaded from disk and the default value AddStakerTx.StartTime() will be used.
- If metadata have version 1, start time will be read from disk.
This is why we don't need a data migration here, which may cause problems with downgrades.
The downside is that it'd be writing the StartTime to disk as soon as the code is deployed, even if Durango is not active. This is an extra int64 written for each staker promoted from pending to current among code deploy and code activation.
There was a problem hiding this comment.
What happens when I upgrade my node to this PR and write codec version v1 and then downgrade to a node that does not support parsing codec version v1?
There was a problem hiding this comment.
@dhrubabasu right! I fixed it by reintroducing the config.Confg and by selecting codec version based on the fork activation. Code will be cleaned up as soon as fork is activated
abi87
force-pushed
the
post_durango_staker_txs
branch
2 times, most recently
from
0deb60a
to
8e8ab33
Compare
abi87
force-pushed
the
post_durango_staker_txs
branch
from
8e8ab33
to
b213740
Compare
StephenButtolph
changed the title
Why this should be merged
Staking will start at a time different from the tx.StartTime specified in each AddStakerTx. Hence we need to explicitly store the start time of each staker to be able to duly rebuild it once we start a node and load data from disk
How this works
Slicing #2175 up
How this was tested
CI