本案例是福建某客户一套待上线系统，操作系统版本为AIX 7.2，集群版本为12.2。几天前2节点GI突然重启，本文将分析其产生的原因。
由于是GI突然重启，那么先看看GI alert日志：
代码语言：javascript复制2021-11-09 21:21:57.050 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:00.066 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:03.070 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:05.093 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:05.151 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:06.081 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:06.093 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:06.100 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:06.105 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:06.109 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:09.091 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:12.097 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:15.100 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:18.105 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:21.109 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:24.113 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:27.118 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:30.122 [GPNPD(7537210)]CRS-2340: Errors occurred while processing received gpnp message.
2021-11-09 21:22:33.126 [OHASD(22348120)]Can't find message for prod[CRS], fac[CRS], and msg id [2119]
2021-11-09 21:22:33.130 [ORAAGENT(17629684)]CRS-5822: Agent '/app/grid/product/12.2.0/grid/bin/oraagent_grid' disconnected from server. Details at (:CRSAGF00117:) {0:2:1632} in /app/grid/base/diag/crs/p1bemdg2/crs/trace/ohasd_oraagent_grid.trc.
2021-11-09 21:22:33.130 [CSSDMONITOR(13042246)]CRS-5822: Agent '/app/grid/product/12.2.0/grid/bin/cssdmonitor_root' disconnected from server. Details at (:CRSAGF00117:) {0:3:21405} in /app/grid/base/diag/crs/p1bemdg2/crs/trace/ohasd_cssdmonitor_root.trc.
2021-11-09 21:22:33.130 [ORAROOTAGENT(38535474)]CRS-5822: Agent '/app/grid/product/12.2.0/grid/bin/orarootagent_root' disconnected from server. Details at (:CRSAGF00117:) {0:1:1420} in /app/grid/base/diag/crs/p1bemdg2/crs/trace/ohasd_orarootagent_root.trc.
2021-11-09 21:22:33.130 [CSSDAGENT(14418484)]CRS-5822: Agent '/app/grid/product/12.2.0/grid/bin/cssdagent_root' disconnected from server. Details at (:CRSAGF00117:) {0:4:21329} in /app/grid/base/diag/crs/p1bemdg2/crs/trace/ohasd_cssdagent_root.trc.可以看到从21点21分开始出现gpnp通信异常，22分33秒时ohasd栈的4个agent全部同时挂掉，这其实给了我们很大的提示，ohasd_orarootagent_root、ohasd_oraagent_grid、ohasd_orarootagent_root、ohasd_cssdagent_root正好是ohasd.bin守护进程启动之后启动的4个agent，它们负责维护(启动、关闭等等)ohasd层面的资源，如果出现同时挂掉的话，那么大概率是ohasd.bin出现了异常crash掉了，所以下一步直接看ohasd的trace文件。
代码语言：javascript复制2021-11-09 21:21:50.060 : OCRUTL:3085: prou_add_log_list_int:proath_new_entry is NULL for 2021-11-09 21:21:50.060: th_process_request: Completed op [13] return [12]
2021-11-09 21:21:50.060 : OCRSRV:3085: th_select_w_f_r: Error processing request [12]
2021-11-09 21:21:50.060 : OCRUTL:3085: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [1040]
2021-11-09 21:21:50.060 : OCRUTL:3085: prou_add_log_list_int:proath_new_entry is NULL for 2021-11-09 21:21:50.060: th_select_w_f_t: Processed request constr [0000000000000069] retval [12]
2021-11-09 21:21:57.032 : OCRUTL:3342: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:3342: proas_get_value: Failure is malloc [12]
OCRAPI:2828: SLOS : SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)
OCRUTL:2828: u_fill_errorbuf: Error Info : [[SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)从gpnpd出现异常的时间点，可以看到ohasd动态分配4120的内存失败，从gipcd.trc也可以看出通信异常的原因也是无法分配内存。
代码语言：javascript复制2021-11-09 21:21:57.034 : OCRCLI:1286: proac_get_value:[SYSTEM.GPnP.profiles.peer.best]: Response message failed [12]
2021-11-09 21:21:57.034 : OCRAPI:1286: a_read: failed to read key [12]
2021-11-09 21:21:57.034 : GPNP:1286: procr_get_value_ext: OLR api procr_get_value_ext failed for key SYSTEM.GPnP.profiles.peer.best
2021-11-09 21:21:57.034 : GPNP:1286: procr_get_value_ext: OLR current boot level : 7
2021-11-09 21:21:57.034 : GPNP:1286: procr_get_value_ext: OLR error code : 12
2021-11-09 21:21:57.034 : GPNP:1286: procr_get_value_ext: OLR error message : PROCL-12: Out of heap memory.
2021-11-09 21:21:57.034 : GPNP:1286: clsgpnp_getProfileEx: [at clsgpnp.c:602] Result: (2) CLSGPNP_MEMORY. (:GPNP00140:) get cache profile failed, will try gpnpd call
2021-11-09 21:21:57.045 : OCRCLI:1286: proac_get_value:[SYSTEM.version.activeversion]: Response message failed [12]
2021-11-09 21:21:57.045 : OCRAPI:1286: a_read: failed to read key [12]
default:1286: Failure 12 in trying to get AV value SYSTEM.version.activeversion
default:1286: procr_get_value error 12 errorbuf : PROCL-12: Out of heap memory.继续往下看ohasd的trace：
代码语言：javascript复制2021-11-09 21:22:18.103 : OCRUTL:2571: prou_add_log_list_int:proath_new_entry is NULL for 2021-11-09 21:22:18.102: th_select_w_f_t: Processed request constr [0000000000000069] retval [0]
OCRAPI:3342: SLOS : SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)
OCRUTL:3342: u_fill_errorbuf: Error Info : [[SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)
]]
2021-11-09 21:22:18.103 : OCRUTL:2828: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:2828: proas_get_value: Failure is malloc [12]
2021-11-09 21:22:18.104 : OCRUTL:3085: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:3085: proas_get_value: Failure is malloc [12]
2021-11-09 21:22:18.104 : OCRSRV:2571: proas_cache_locked: Value not cached for key [SYSTEM.network] : retval [84]
OCRAPI:3342: SLOS : SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)
OCRUTL:3342: u_fill_errorbuf: Error Info : [[SLOS: cat=-2, opn=malloc, dep=12, loc=addgrpnm1
unable to add group (system)
]]
2021-11-09 21:22:18.105 : OCRUTL:3085: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:3085: proas_get_value: Failure is malloc [12]
2021-11-09 21:22:18.107 : OCRUTL:2571: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [280]
2021-11-09 21:22:18.107 : OCRRAW:2571: prcreatei: could not allocate memory 3
...
2021-11-09 21:22:33.126 : OCRUTL:2571: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:2571: proas_get_value: Failure is malloc [12]
2021-11-09 21:22:33.126 :GIPCXCPT:2057: gipclibMalloc: failed to allocate 40960 bytes, cowork 11234c978, ret gipcretOutOfMemory (28)
2021-11-09 21:22:33.126 : OCRUTL:2828: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [4120]
OCRSRV:2828: proas_get_value: Failure is malloc [12]
2021-11-09 21:22:33.126 :GIPCXCPT:2057: gipcWaitF [prom_select : prom.c : 4591]: EXCEPTION[ ret gipcretOutOfMemory (28) ] failed to wait on obj 112592dd0 [000000000000002a] { gipcContainer : numObj 11, wobj 1125933b0, objFlags 0x0 }, reqList 11234d420, nreq 1, creq 11234d4e8 timeout INFINITE, flags 0x0
2021-11-09 21:22:33.126 : OCRUTL:2828: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [1040]
2021-11-09 21:22:33.126 : OCRMSG:2057: prom_select: Failed to wait on container [28]
2021-11-09 21:22:33.126 : OCRUTL:2828: prou_add_log_list_int:proath_new_entry is NULL for 2021-11-09 21:22:33.126: prom_send: Before gipcSend
2021-11-09 21:22:33.126 : OCRMSG:2057: GIPC error [28] msg [gipcretOutOfMemory]
2021-11-09 21:22:33.126 : OCRUTL:2571: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [1040]
2021-11-09 21:22:33.126 : OCRUTL:2571: prou_add_log_list_int:proath_new_entry is NULL for 2021-11-09 21:22:33.126: th_select_w_f_t: Processed request constr [0000000000000069] retval [0]
2021-11-09 21:22:33.126 : OCRUTL:2057: u_allocm: metactx [111a7c2d0], ptr returned is NULL, size to lmmcalloc [2088]
2021-11-09 21:22:33.126 : OCRUTL:2057: u_insert_error: Error allocating memory for hashelement.
OCRSRV:2057: th_select: Failed to select [24]
OCRSRV:2057: th_select: Can no longer process local connections. Setting OCR context to invalid
2021-11-09 21:22:33.126 : OCRAPI:2057: procr_ctx_set_invalid: ctx is in state [10].
2021-11-09 21:22:33.126 : OCRAPI:2057: procr_ctx_set_invalid: ctx set to invalid
2021-11-09 21:22:33.126 : OCRAPI:2057: procr_ctx_set_invalid: Aborting...
Trace file /app/grid/base/diag/crs/p1bemdg2/crs/trace/ohasd.trc
Oracle Database 12c Clusterware Release 12.2.0.1.0 - Production Copyright 1996, 2016 Oracle. All rights reserved.
CLSB:1: Argument count (argc) for this daemon is 2
CLSB:1: Argument 0 is: /app/grid/product/12.2.0/grid/bin/ohasd.bin
CLSB:1: Argument 1 is: restart可以看到大量的内存分配失败之后，ohasd.bin被abort了，这就是导致GI重启的原因。由于ohasd.bin是root来维护的，所以查看一下root的ulimit：
这里可以看到大量错误的设置，包括导致此次故障的memory limit。
墨天轮原文链接：https://www.modb.pro/db/174953?sjhy（复制链接至浏览器或点击文末阅读原文查看)
关于作者李翔宇，云和恩墨西区交付技术顾问，长期服务移动运营商行业客户，熟悉Oracle性能优化，故障诊断，特殊恢复。
END
推荐阅读：331页！2021年度数据库技术年刊
推荐下载：2021数据技术嘉年华视频回放及PPT下载
2021数据技术嘉年华50余个PPT下载、视频回放已上传墨天轮平台，可在“数据和云”公众号回复关键词“2021DTC”获得！
你知道吗？我们的视频号里已经发布了很多精彩的内容，快去看看吧！↓↓↓
点击下图查看更多 ↓
云和恩墨大讲堂 | 一个分享交流的地方
长按，识别二维码，加入万人交流社群
请备注：云和恩墨大讲堂
点个“在看”
你的喜欢会被看到❤
