PowerStore: Node May Reboot Unexpectedly from a Linux Driver Memory Leak

摘要: A node may reset unexpectedly due to a memory leak within kernel space for SCSI Target (SCST) Report LUNs response.

本文适用于 本文不适用于 本文并非针对某种特定的产品。 本文并非包含所有产品版本。
查看其他资源

症状

The system reports excessive messages such as:
svc_journalctl | grep "REPORT LUNS"

Apr x 01:00:31 xxxxx-A kernel: scst: REPORT LUNS got, target='xx:xx:xx:xx:xx:xx:xx:xx', initiator='xx:xx:xx:xx:xx:xx:xx:xx'
Apr x 01:00:31 xxxxx-B kernel: scst: REPORT LUNS got, target='xx:xx:xx:xx:xx:xx:xx:xx', initiator='xx:xx:xx:xx:xx:xx:xx:xx'
Apr x 01:00:33 xxxxx-A kernel: scst: REPORT LUNS got, target='xx:xx:xx:xx:xx:xx:xx:xx', initiator='xx:xx:xx:xx:xx:xx:xx:xx'
Apr x 01:00:34 xxxxx-B kernel: scst: REPORT LUNS got, target='xx:xx:xx:xx:xx:xx:xx:xx', initiator='xx:xx:xx:xx:xx:xx:xx:xx'
Apr x 01:00:34 xxxxx-A kernel: scst: REPORT LUNS got, target='xx:xx:xx:xx:xx:xx:xx:xx', initiator='xx:xx:xx:xx:xx:xx:xx:xx'



A useful command to check the number of these messages is, for example:
 
svc_journalctl --since="today" | grep -c "REPORT LUNS"
44835


Indications of starvation may start coming up in the logs:
 
svc_journalctl --since="today" | grep -i starved

Apr x 01:47:58.501672 xxxx-A xtremapp[22524]: Apr x 01:47:58.500331 M [log_id:22772][3840(3891 nb_truck_0_sym   0x7fa0ffc20d00)]sym_ham_nbt_timer:1284: ====+ HAM: SYM NBT was starved for 550906 us (the threshold is 500000) #STARV
Apr x 03:14:00.770872 xxxx-A xtremapp[22524]: Apr x 03:14:00.745703 M [log_id:22772][3840(3891 nb_truck_0_sym   0x7fa0ffc20d00)]sym_ham_nbt_timer:1284: ====+ HAM: SYM NBT was starved for 518854 us (the threshold is 500000) #STARV
Apr x 04:35:39.509945 xxxx-A xtremapp[22524]: Apr x 04:35:39.507440 M [log_id:22772][3840(3891 nb_truck_0_sym   0x7fa0ffc20d00)]sym_ham_nbt_timer:1284: ====+ HAM: SYM NBT was starved for 523371 us (the threshold is 500000) #STARV


A dump file is created for this event but it is not required for the Root Cause Analysis (RCA). See svc_dc list_dumps.

原因

This issue happens due to a memory leak within the kernel space for SCST due to excessive "Report LUN" commands from ESXi. This may cause a delay in keepalive communication between nodes triggering a node reset.

解决方案

This issue is resolved in PowerStoreOS 1.0.4.0.5.006 (V1 SP4.1).

The fix is listed in the release notes under MDT-261192: A node may reboot unexpectedly, due to a Linux driver memory leak.

受影响的产品

PowerStore, PowerStore 1000X, PowerStore 1000T, PowerStore 3000X, PowerStore 3000T, PowerStore 5000X, PowerStore 5000T, PowerStore 7000X, PowerStore 7000T, PowerStore 9000T

产品

PowerStore 9000X
文章属性
文章编号: 000185532
文章类型: Solution
上次修改时间: 04 11月 2021
版本:  5
从其他戴尔用户那里查找问题的答案
支持服务
检查您的设备是否在支持服务涵盖的范围内。