F country a customer site finds several hosts connected to storage cannot access application normally
4 BMPAS single machine: HP6600
1 suit of BMPDB dual machine: HP7640
6 machines is connected to a set of S2300R disk array (two arrays with relationship of concatenation)
Oracle edition: Oracle11g
Problem symptom description:
BMPAS1A: unable to log on, when log on with MP port, cannot get in CO
BMPAS2A: able to log on, but when bdf run, there is phenomenon like in attachment figure. Su ? bmp user, phenomenon is like in attachment figure. syslog_BMPAS2A.log
BMPAS3A：able to log on, but when bdf run, there is phenomenon like in attachment figure. syslog_BMPAS3A.log
BMPAS4A：it looks like normal. Corresponding log: syslog_BMPAS4A.log
BMPDB1A: unable to log on, system start excerptly
BMPDB1B: log correspond to syslog_BMPASDB_standby.log
1. Reboot storage controller.
2. Change LUN Cache write policy from write through to write back, avoid that this problem happen again.
3. Recommend to upgrade controller software, avoid that this problem happen again. Upgrade from V100R001C02SPC020 (1.02.01.220.T01) to V100R001C02SPC025 (1.02.01.223.T06).
France SFR program S2300E host cannot access storage normally. Analyze array log, cache is loaded with IO, the LUN with IO is write through and no mirror. The main cause of this problem is: controller software Cache write through process has BUG, this BUG would cause IO loaded to Cache and do not return outcome to host in a low probability, which cause IO timeout failure.