Faster node initialization #1756
Conversation
Cap at Environment.ProcessorCount. Start loading biggest index files first.
|
@megakid thanks very much! It seems that this recently added test is failing: I've done a quick test and could see the following output: It's strange, looks like the chunk corruption was not detected. It sounds plausible that the aggregate exception may not have been thrown properly, what do you think? (or it could be an issue with the test) |
Removed `OpenOld` method that was unused.
Hi @shaan1337 - I found the bug which was an off-by-one error in the chunk numbering during the hash check code. This part of the method is a direct copy from the original code so not sure where the off-by-one error came in - My only explanation is that maybe this code was fixed in the Fixed now 👍 |
|
@megakid thanks very much. that makes sense, it was fixed in this PR: #1653 It's strange that it didn't pick up the change after the rebase - I guess that something went wrong there. (maybe |
| } | ||
| catch (Exception exc) | ||
| { | ||
| var msg = string.Format("Verification of chunk {0} failed, terminating server...", chunk); | ||
| Log.FatalException(exc, msg); | ||
| throw new CorruptDatabaseException(msg, exc); | ||
| Application.Exit(ExitCode.Error, msg); |
There was a problem hiding this comment.
I've noticed that this line has been reverted here. It's better to keep it this way. The initial plan was to make this testable by catching the exception but right now it cannot be caught from the main thread since it's thrown from a threadpool thread.
ChrisChinchilla
removed
the
area/documentation
Had to open a new PR for this
masterrebase of #1642