jja725 comments

Results 33 comments of


                                            jja725

Optimizing reading and writing performance

mark

After distributedLoad , the job master Process GC CPU usage is too high, and lasts a long time.

@LKR1996 Currently it's a bit hard to tell why we have this issue. Do you mind uploading the log in that time range? Also, a heap dump of job master...

After distributedLoad , the job master Process GC CPU usage is too high, and lasts a long time.

@LKR1996 do you mind doing me a favor and testing version 2.7? One of the changes we have in 2.8 is adding more metrics to distributedLoad, so we can sort...

Implement zero copy local cache interface

> hello @jja725, just wanna address your review comment, I don't know why I cannot reply it. I had tried to use PooledByteBufAllocator below, but it's still calling allocateDirect somehow,...

Add microbenchmarks for multiple implementations of BlockStore

It's surprising that monoBlockStoreReadLocal performance is so bad...

Alluxio files can be deleted during persisting, but the persist job was not interrupted

yeah, I think that's a current problem with how job service handles jobs. And we are planning on a new version job service which would include a more stable Persist...

Fix NPE during DelegatingBlockReader initialization with debug logging

hi, thanks for the contribution! do you mind put this into master-2.x? We can later on cherry-pick this to 2.9.1

there are many SIGSEGV message in the strace log of the alluxio fuse process

@LuQQiu @jiacheliu3 Do you have any idea on this issue?

Support redis as an UFS of Alluxio

Will close it for now, feel free to reopen it and contact us if this is a strong request.

Fix data corrupted issue when loading from UFS

@yuzhu I think these tests are still broken. @beinan I can take over and take a look if you want