Multipath device is rejected and subsequently not created; size discard and domap failed to add paths
背景
部分多路径磁盘无法聚合multipath -v2有如下提示信息
reject: 363c9d56100e06a30eed2649d00000075 undef HUAWEI ,XSG1
size=200G features='0' hwhandler='0' wp=undef
`-+- policy='service-time 0' prio=-1 status=undef
|- 7:0:4:8 sdaq 66:160 undef ready running
`- 7:0:3:8 sdl 8:176 undef ready running
Aug 25 02:24:55 | sdar: alua not supported
Aug 25 02:24:55 | sdm: alua not supported
reject: 363c9d56100e06a30eed2655600000076 undef HUAWEI ,XSG1
size=200G features='0' hwhandler='0' wp=undef
`-+- policy='service-time 0' prio=-1 status=undef
|- 7:0:4:9 sdar 66:176 undef ready running
`- 7:0:3:9 sdm 8:192 undef ready running
事情是这样的，有一套ORACLE RAC 数据库之前使用的SAN存储准备下线，计划采用新的华为存储替换。将主机HBA卡一个端口的光纤从原来老存储光交拔下插到新存储所在光交上，存储上划好LUN，主机上扫盘，磁盘正常识别，并被多路径以默认规则聚合。问题是当我在配置文件/etc/multipath.conf为这些LUN添加好别名规则后，刷新或重启multipath都无法聚合新的盘，即使还原了配置文件也不行，老存储的LUN可以正常聚合。multipath -v2可以看到有reject.....的信息。查询redhat资料库了解到，应该是新LUN size 属性和原来的LUN不一致导致的。重启主机问题就解决了。如下是官方解析。
Multipath device is rejected and subsequently not created; size discard and domap failed to add paths
https://access.redhat.com/solutions/4919741)
环境
- Red Hat Enterprise Linux 5
- Red Hat Enterprise Linux 6
- Red Hat Enterprise Linux 7
- Device-mapper-multipath
问题
Multipathdoes not create amultipathdevice when the nativempioandSANconfiguration is such that a device should be created.
决议
Rescan scsi devices, or delete and rescan scsi devices, or a reboot of the system should resolve this problem:
# echo 1 > /sys/block/sdX/device/rescan
or
# echo 1 > /sys/block/sdX/device/delete
# rescan-scsi-bus.sh -a <<<-------requires sg3_utils to be installed
Rescan for multipath devices:
# multipath v2
# multipath -ll
Verify the previously rejected multipath device has now been created.
根源
The sd device(s) in sysfs contains stale size attributes, resulting in attributes that do not match the other sd devices associated with the LUN. When multipath checks the sd devices associated with a SAN LUN, and it detects size inconsistencies, it will discard the sd devices associated with the LUN, reject the LUN, and generate domap failed logs for the underlying sd devices.
From the code:
for (i = k + 1; i < VECTOR_SIZE(pathvec); i++) {
pp2 = VECTOR_SLOT(pathvec, i);
if (strcmp(pp1->wwid, pp2->wwid))
continue;
if (!pp2->size)
continue;
if (pp2->size != mpp->size) {
/*
* ouch, avoid feeding that to the DM
*/
condlog(0, "%s: size %llu, expected %llu. "
"Discard", pp2->dev_t, pp2->size,
mpp->size);
mpp->action = ACT_REJECT;
}
if (pp2->priority == PRIO_UNDEF)
mpp->action = ACT_REJECT;
}
诊断步骤
Check for reject and discards in the multipath output:
# multipath -v2
Mar 23 11:56:17 | 8:16: size 16780800, expected 134219520. Discard
Mar 23 11:56:17 | 8:26: size 16780800, expected 134219520. Discard
Mar 23 11:56:17 | 8:114: size 16780800, expected 134219520. Discard
Mar 23 11:56:17 | DM message failed [queue_if_no_path]
reject: mpathab (360000000000012345678901234567890) undef EMC ,SYMMETRIX
If you look at the -v4 output, or check the size attributes in sysfs, at least one of the devices reports a size that does not match the others:
# multipath -v4
[...]
Mar 23 15:51:20 | sdab: size = 16780800
Mar 23 15:51:20 | sdbd: size = 134219520 <<<---
Mar 23 15:51:20 | sdca: size = 16780800
Mar 23 15:51:20 | sddi: size = 16780800
and
# for i in ab bd ca di; do cat /sys/block/sd$i/size; done
16780800
134219520 <<<---
16780800
16780800
