Attender: Guohao, Minyu, Yiyang, XiChen, Hongbing, Guangbao, Hualong,
Yuanben, Sammi
Agenda:
1.
Guohao shared the experience about how to migrate Container Schema V2
data to Schema V3, the problem they met and how it’s resolved. In their
last try,, it took around 15 min to migrate a disk volume(8TB, 60%
utilization) data from V2 to V3.
2.
Hongbin and Guangbao noticed that after a very high data ingestion
workload, the Datanode instance will occupy almost all the memory on this
server, mostly were off-heap memory usages. And after the workload stopped,
the memory consumption of the Datanode instance didn’t shrink obviously.
There is potentially an off-heap memory leak here. They will continue to
monitor the service and collect more data/info for further investigation.
3.
1.4.0 Release
1.
RATIS 3.0.0 enabling in Ozone is ongoing. There are some snapshot
related CI failures and need fixing.
2.
Yiyang has filtered all open blockers and critical issues which
targets 1.4.0. Most of them are snapshot improvement and certificate
rotation patches.
4.
For “Missing” containers shown on recon, Guangbao would like to know how
to identify if there is real data missing or not. One way is to search this
container ID in OM DB, to see if there is any file/object whose block
location contains the container ID. For the coming 1.4.0 release, there are
new features on Recon which help to further analyze these “Missing”
container cases.
