[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[dm-devel] system freeze/console flood when unplugging one path of multipath disk array



Hello,

I've got a Sun 3510 disk array and a dual-channel Qlogic FC HBA, in a Sun X4100 under RHEL4 U4, each path connected to the array over a qlogic 5600 switch.

The devices are seen correctly with multipath -l as far as I can judge.


When I disable one of the paths during a stress test (in this case, an fsck -y), the system "freezes". Actually, it just sends enough (thousands) error messages of the style

end_request: I/O error, dev sdb, sector 279446024
Buffer I/O error on device sdb1, logical block 279445961
lost page write due to I/O error on sdb1
SCSI error : <1 0 0 1> return code = 0x10000
end_request: I/O error, dev sdb, sector 279446025
Buffer I/O error on device sdb1, logical block 279445962
lost page write due to I/O error on sdb1
SCSI error : <1 0 0 1> return code = 0x10000
end_request: I/O error, dev sdb, sector 279446026
Buffer I/O error on device sdb1, logical block 279445963
lost page write due to I/O error on sdb1
SCSI error : <1 0 0 1> return code = 0x10000
end_request: I/O error, dev sdb, sector 279446027
Buffer I/O error on device sdb1, logical block 279445964
lost page write due to I/O error on sdb1
SCSI error : <1 0 0 1> return code = 0x10000


to the serial console to make it appear completely stuck for hours.

The fsck does not finish, BTW, it restarts from the beginning as soon as I re-enable the port in the switch.

root afs26 ~]# multipath -l
mpath2 (3600c0ff0000000000a792a1887b34100)
[size=136 GB][features="1 queue_if_no_path"][hwhandler="0"]
\_ round-robin 0 [active]
 \_ 1:0:0:3  sdc 8:32  [active][ready]
 \_ 2:0:0:3  sdi 8:128 [active][ready]

mpath1 (3600c0ff0000000000a792a0aac9b4200)
[size=136 GB][features="1 queue_if_no_path"][hwhandler="0"]
\_ round-robin 0 [active]
 \_ 1:0:0:1  sdb 8:16  [active][ready]
 \_ 2:0:0:1  sdh 8:112 [active][ready]
....

/etc/multipath.conf just contains
defaults {
        default_features "1 queue_if_no_path"
        user_friendly_names yes

}

besides the usual blacklist.

Any ideas?

--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rainer Toebbicke
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland
Phone: +41 22 767 8985       Fax: +41 22 767 7155


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]