[MESOS-1014] Log truncation takes a long time during catch-up if the initial position is very large

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.18.0
• Fix Version/s: 0.17.0
• Assignee: Jie Yu
• Reporter: Jie Yu
• Created: 18/Feb/14 22:02
• Updated: 24/Feb/14 19:40
• Resolved: 19/Feb/14 18:51

Description

in src/log/leveldb.cpp

The while loop takes a long time if action.truncate().to() is very large.

Try<Nothing> LevelDBStorage::persist(const Action& action)
{
   ...
       leveldb::WriteBatch batch;

       // Add positions up to (but excluding) the truncate position to
       // the batch starting at the first position still in leveldb.
       uint64_t index = 0;
       while ((first + index) < action.truncate().to()) {
          batch.Delete(encode(first + index));
          index++;
       }
  ...
}