Esxi loganalysis :
In log ,large number of the record as below was printed :
cpu105:2099446)ScsiDeviceIO: 3414:Cmd(0x45a57aba6a00) 0x42, CmdSN 0x1eb18 from world 2135471 to dev"naa.6488eef100878a603447c0550000000c" failed H:0x0 D:0x8 P:0x0Invalid sense data: 0x0 0x0 0x0.
cpu91:2099445)NMP:nmp_ThrottleLogForDevice:3802: Cmd 0x42 (0x45a166393600, 2135467) to dev"naa.6488eef100878a603447c0550000000c" on path"vmhba3:C0:T26:L2" Failed: H:0x0 D:0x8 P:0x0 Invalid sense data: 0x00x0 0x0. Act:NONE
Cmd 0x42means unmap command , it was failed due to “H:0x0 D:0x8 P:0x0” , this code was returnedfrom storage , it means that storage was busy .
Storagelog analysis:
According to the time when theesxi log show errors . we found that there are many error records about unmapspeed was limited by storage .
2.Check the LUN print information.It is found that the I/O block size is 16 MB, and the storage system keepsreturning busy to the host for retry. This indicates that the unmap speedlimitation takes effect.
According to the speed limitationstatistics at the log collection time point, the maximum unmap traffic ofcontroller B is about 230 MB (the maximum value in 30 seconds), and the averagetraffic is 400 KB (the statistics are collected in 5 seconds). (this can bequeried in qos_info under other folder .)
【Root Cause】 : Afterthe unmap speed limitation is triggered on the storage side, the storagereturns busy to the host.
【Solution Description】 : Whenissue occur next time , customer can execute “overload_ctrl show flowctrl_info”under debug mode of storage (enter developer mode then execute “debug” to enterdebug mode). Try more times to ensure the value queried is mostly possiblecorrect
