This bug is fixed now. But how often will we change the HBase client version in Phoenix pom? Will it be better to have a mitigation in Phoenix code also. Simple change in ScanUtil.java to have a isSmall() true check before calling newScan.setSmall(scan.isSmall());
This will create a big perf issue specially for cloud based storage where buffered reads are very key. I had a big perf regression for a WASB backed HBase cluster. Anoop
