Bug 13033
Summary: | bonding leads to iscsi initiator error | ||
---|---|---|---|
Product: | Networking | Reporter: | zhanghj (zhanghj_2000) |
Component: | IPV4 | Assignee: | Stephen Hemminger (stephen) |
Status: | CLOSED OBSOLETE | ||
Severity: | high | CC: | akpm, alan, zhanghj_2000, zhlqcn |
Priority: | P1 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.24.7 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
zhanghj
2009-04-07 02:48:46 UTC
i change the bonding mode to mode=0(balance-rr) and mode=5(balance-tlb), have the same problem, but change to mode=2(balance-xor) the iometer can run normally. I dicussed this with Mike Christie <michaelc@cs.wisc.edu> and Jay Vosburgh <fubar@us.ibm.com>. Mike said: I think they are using a microsoft initiator/host with a non-upstream iscsi target/server, so I am not 100% sure. I do not work on either project so I am just guessing below :) From the target logs in the bugzilla, it looks like there was probably disruption with the connection or a command took too long. Then the windows initiator probably tried to run some error handling (lun reset). From there it looks like the target and initiator could not agree on what was the proper next step (initiator might have sent data when it should not have). The initiator eventually gave up and removed the disks thinking they were bad or because it could not figure out anything else to try. The bug report should go to iscsitarget-devel@lists.sourceforge.net They maintain the IET iscsi target, and can better debug problems against the microsoft initiatior since they have people that know how that works. |