-
Notifications
You must be signed in to change notification settings - Fork 5.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Change osd journal padding mode. #5877
Conversation
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
The reasons are align_bl don't need protect of write_lock and align_bl may take long time(do memcopy data). Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Current, because the requirement of DirectIo, the padding mode is adding padding data for every entry. In fact, we fetch a couple of entries as a system write as possible as can. So we only make sure system-write content met the requirement of DirectIO. We adding a padding entry at the end of system-write content(for most case, this include a couple of entries). Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
@liewegas . Ping! |
@majianpeng , I think it is a good strategy for hdd. But In my test, the write journal thread is bottleneck in fast equipment because of the crc and rebuild logic, such as pcie ssd. You could get more detail information from ceph-devel mailist. |
@majianpeng , @liewegas , I think this is make sense for some type of equipments. Please review it #6856. |
I'm very nervous about making this sort of change unless there is a demonstrated and significant performance improvement. Otherwise we risk introducing new bugs in critical code for little or no gain. |
Now for every entry, we pad data if need. But for most case, we fetch more entries as a system-write. So we can add padding data at the end of system-write content if need. This can reduce much padding data for small write.