環境介紹：
| 服務器 | vShere 6 虛擬機 |
| 操作系統 | OEL 7.0 with Kernal 3.8.13-35.3.1.el7uek.x86_64 |
| 數據庫 | Oracle RAC 12.2.0.1 |
| 節點數 | 3 |
現象：
首先是使用oratop命令的時候只有在節點1執行成功，其他節點出現ORA-12805的錯誤。
並且節點1上的Top Waits Events很不正常:
后來發現只有節點1可以查詢GV$視圖，其余兩個節點查詢時很慢並報錯如下：
SQL> select count(*) from gv$instance;
select count(*) from gv$instance
*
ERROR at line 1:
ORA-12805: parallel query server died unexpectedly
解決方案：
MOS上搜了一下，發現有同樣的現象描述Doc ID 2014298.1，同時下面的這個文章里有對應的Patch修復：
NOTE:1347455.1 - OERR: ORA 12805 parallel query server died unexpectedly
| IIII | 27223075 | 12.1.0.2.190115, 19.1 | Wait for 'PX Deq: Join Ack' when no active QC but PPA* slaves show as busy |
跟上面的Wait Event相同。
摘抄MOS上的描述：
Description
This bug is only relevant when using Real Application Clusters (RAC) and Parallel Query (PQO)
Hang issue during slave join period and have to kill slave process to end the
hang.
Rediscovery:
A PPA* slave that is busy is thought to be available by a QC and
is asked to join, but the slave cannot report back to the QC, so
it hangs and has to be killed.
You may run the following queries and see no rows returned from
the first one, and PPA* slaves returned from the second one:
1. No Query slaves running.
select
decode(px.qcinst_id,NULL,username,
' - '||lower(substr(s.program,length(s.program)-4,4) ) ) "Username",
decode(px.qcinst_id,NULL, 'QC', '(Slave)') "QC/Slave" ,
to_char( px.server_set) "Slave Set",
to_char(s.sid) "SID",
decode(px.qcinst_id, NULL ,to_char(s.sid) ,px.qcsid) "QC SID",
px.req_degree "Requested DOP",
px.degree "Actual DOP"
from
v$px_session px,
v$session s
where
px.sid=s.sid (+)
and
px.serial#=s.serial#
order by 5 , 1 desc
/
no rows selected
...
2. But PPA4/PPA5/PPA6/PPA7 are busy.
SQL> select SLAVE_NAME from V$PQ_SLAVE where STATUS='BUSY';
SLAV
----
PPA4
PPA5
PPA6
PPA7
Additional Info:
Any parallel slave process (p00*) can be affected by this bug and is not constrained to PPA* slaves.
Workaround
None.