Re: bin/56728: iscsi redundancy doesn't work

[mlelstv%serpens.de@localhost (Michael van Elst)]

The following reply was made to PR bin/56728; it has been noted by GNATS.

From: mlelstv%serpens.de@localhost (Michael van Elst)
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: bin/56728: iscsi redundancy doesn't work
Date: Sat, 30 Apr 2022 10:02:23 -0000 (UTC)

 6bone%6bone.informatik.uni-leipzig.de@localhost writes:
 
 >Details on problem 1: If you use iscsictl to open and close connections to 
 >the iscsi target, the kernel crashes. iscsi devices are not mounted at 
 >this time. Here is a kernel dump:
 
 >[ 40282.903575] trap type 6 code 0x2 rip 0xffffffff8022d84c cs 0x8 rflags 
 >0x10246 cr2 0xffffa080679011c0 ilevel 0 rsp 0xffffa08397034f08
 >[ 40282.903575] curlwp 0xffffecdb7be4db00 pid 0.581 lowest kstack 
 >0xffffa083970302c0
 >[ 40282.903575] panic: trap
 >[ 40282.903575] cpu0: Begin traceback...
 >[ 40282.903575] vpanic() at netbsd:vpanic+0x183
 >[ 40282.903575] panic() at netbsd:panic+0x3c
 >[ 40282.903575] trap() at netbsd:trap+0xb27
 >[ 40282.903575] --- trap (number 6) ---
 >[ 40282.903575] mutex_enter() at netbsd:mutex_enter+0xc
 >[ 40282.903575] send_nop_out() at iscsi:send_nop_out+0x133
 >[ 40282.903575] connection_timeout() at iscsi:connection_timeout+0x4d
 >[ 40282.903575] iscsi_cleanup_thread() at iscsi:iscsi_cleanup_thread+0x7b2
 >[ 40282.903575] cpu0: End traceback...
 >[ 40282.903575] dumping to dev 4,1 (offset=22227071, size=12581616):
 >[ 40282.903575] dump <7>lagg0: link state DOWN (was UP)
 
 
 If that's reproducable, please provide a log with ISCSI_DEBUG.
 
 
 
 >Problem 2 details: We are using a NetApp iscsi cluster. Multiple Netapp 
 >controllers are used for redundancy. These provide the same device (iqn). 
 >If you establish a session to the second controller, NetBSD does not 
 >recognize that it is the same iqn. NetBSD thinks it's a new device:
 
 It is a new device. Linux supports multipathing (in the device mapper
 driver) that detects such duplicates, handles the redundancy and
 provides a virtual disk.
 
 Unfortunately our device mapper still lacks multipathing support
 (and also mirror support, but that's another story).
 
 
 
 
 >Problem 3: the man page for iscsictl contains errors. For example, the 
 >add_connection command requires an -I parameter that is not described. The 
 >meaning of -m is not described. There seem to be errors with other 
 >commands as well.
 
 Yes. The man page is neither correct nor complete.
 
 
 
 Greetings,