You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When the consumer WAL log encounters the latest Sequence Number, the status of the iterator should be OK, not "tryagain".
Reason
This kind of consumption WAL log speed catch up with the latest Sequence scene. If it is returned to "TRYAGAIN", according to semantics, a new WAL iterator must be allocated (because it cannot be distinguished by key, or other situations, so), when WAL When the file is large, this will lead to a large part of the WAL log from scanning, plus CRC32C computing, resulting in a large delay in the production environment.
Actual behavior
When the consumer WAL log encounters the latest serial number, the state of the iterator returns "TRYAGAIN".
Steps to reproduce the behavior
The text was updated successfully, but these errors were encountered:
threadfly
pushed a commit
to threadfly/rocksdb
that referenced
this issue
May 5, 2022
Expected behavior
When the consumer WAL log encounters the latest Sequence Number, the status of the iterator should be OK, not "tryagain".
Reason
This kind of consumption WAL log speed catch up with the latest Sequence scene. If it is returned to "TRYAGAIN", according to semantics, a new WAL iterator must be allocated (because it cannot be distinguished by key, or other situations, so), when WAL When the file is large, this will lead to a large part of the WAL log from scanning, plus CRC32C computing, resulting in a large delay in the production environment.
Actual behavior
When the consumer WAL log encounters the latest serial number, the state of the iterator returns "TRYAGAIN".
Steps to reproduce the behavior
The text was updated successfully, but these errors were encountered: