Qian Bin

The buffer pool is now mainly for reusing block read/write buffers. After #367 merged, the baseline does not work well as wish. Last two commits simplify BufferPool and highly increase...

> Could you check this allocator? https://github.com/xtaci/smux/blob/master/alloc.go It's limited to 64KB?

@rjl493456442 you're right. The manifest file is correctly written without duplicated records. The problem is about table reference count. Here https://github.com/syndtr/goleveldb/blob/126854af5e6d8295ef8e8bee3040dd8380ae72e8/leveldb/session_util.go#L420 if we pass the newly committed `rec` object, it...

the last commit checks the limit on db open. @StephenButtolph it's reverted to use for loop to double the limit.

@StephenButtolph Yes, the edge case can happen. > The way I'd have thought to handle this is just doubling s.maxManifestFileSize every time the file gets re-written, regardless of the new...

> @qianbin > please merge my PR I'm sorry that i'v no privilege to merge PR.

You are right. There's no way to get node version. However, the version of REST API can be found in the field `X-Thorest-Ver` of response headers. ``` curl -I http://localhost:8669/blocks/best...

> > Can I set threads limit? I think it too much for 4-8 vCPU. > > ``` > # ps -e -T | grep 19062| wc -l > 217...

how long did it last? It's most likely the state pruner was running, if just for several minutes. you can filter out logs with keyword 'pruner'.

so it's nothing related to pruner. wondering how is the disk load if all http requests are stopped.