NOTE: Failed voting file relocation on diskgroup

After Oracle support performs maintenance for Exadata Cloud, there are many errors from ASM alert log.

There have been many discussions whether it’s a BUG or if alert can be ignored.

Unfortunately, BUG does not fit and cannot be ignored.

The simple solution is to check for SUCCESS after Failed as shown below.

egrep -n 'NOTE: Failed voting file relocation on diskgroup|SUCCESS: refreshed membership' /u01/app/grid/diag/asm/+asm/+ASM1/trace/alert_+ASM1.log|grep DATAC1
1937:SUCCESS: refreshed membership for 1/0xacd2391b (DATAC1)
1972:SUCCESS: refreshed membership for 1/0xacd2391b (DATAC1)
4244:NOTE: Failed voting file relocation on diskgroup DATAC1
4250:SUCCESS: refreshed membership for 1/0x8f4036a8 (DATAC1)
9876:NOTE: Failed voting file relocation on diskgroup DATAC1
9881:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
10130:NOTE: Failed voting file relocation on diskgroup DATAC1
10135:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
11112:NOTE: Failed voting file relocation on diskgroup DATAC1
11425:NOTE: Failed voting file relocation on diskgroup DATAC1
11441:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
12410:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
13318:NOTE: Failed voting file relocation on diskgroup DATAC1
13717:NOTE: Failed voting file relocation on diskgroup DATAC1
13733:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
14703:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
15566:NOTE: Failed voting file relocation on diskgroup DATAC1
15865:NOTE: Failed voting file relocation on diskgroup DATAC1
15881:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)
16836:SUCCESS: refreshed membership for 1/0x8f4038d9 (DATAC1)

Want more comfort?

$ crsctl query css votedisk
##  STATE    File Universal Id                File Name Disk group
--  -----    -----------------                --------- ---------
 1. ONLINE   1b45f107ff974f1abf34ff7b005eaff1 (/dev/exadata_quorum/QD_DATAC1_QM828NAKVNXA1) [DATAC1]
 2. ONLINE   b3607b245b984f7ebfe5188c0775c44e (/dev/exadata_quorum/QD_DATAC1_QM738NAKVNXA2) [DATAC1]
 3. ONLINE   5d26163d434a4fb6bf2cb173bae3cae1 (o/192.168.136.14;192.168.136.15/DATAC1_CD_05_phx302307exdcl07) [DATAC1]
 4. ONLINE   3e1661086fed4f8bbf080db321282b23 (o/192.168.136.16;192.168.136.17/DATAC1_CD_04_phx302307exdcl08) [DATAC1]
 5. ONLINE   edac9a7822624fe4bfd59eb357d55d95 (o/192.168.136.18;192.168.136.19/DATAC1_CD_04_phx302307exdcl09) [DATAC1]
Located 5 voting disk(s).

# ocrcheck
Status of Oracle Cluster Registry is as follows :
         Version                  :          4
         Total space (kbytes)     :     491684
         Used space (kbytes)      :      90748
         Available space (kbytes) :     400936
         ID                       : 1880339001
         Device/File Name         :    +DATAC1
                                    Device/File integrity check succeeded
                                    Device/File not configured
                                    Device/File not configured
                                    Device/File not configured
                                    Device/File not configured
         Cluster registry integrity check succeeded
         Logical corruption check succeeded

Advertisement

Monitor Linux Host Restart

The application is not RAC-aware and cannot handle ORA-3113, ORA-25402, or ORA-25409 properly.

Hence, there is requirement to notify the application team to restart the application when database server is restarted.

Initial implementation to monitor reboot was to use cronjob from oracle running every 5m to detect server restart.

While the implementation is effective, it’s not efficient. This was my first attempt.

The script detects if server was restarted X seconds ago by checking /proc/uptime.

If uptime is less than X seconds, then send notification server was restarted.

Here is high level example:

### Scripts accept paramenter with values for seconds
$ /home/oracle/scripts/last_reboot.sh
/home/oracle/scripts/last_reboot.sh: line 10: 1: ---> USAGE: /home/oracle/scripts/last_reboot.sh [in seconds]

### The heart of the script is to check /proc/uptime in seconds
$ egrep -o '^[0-9]+' /proc/uptime
2132607

### Scheduled cron tab to run every 5 minute to determine if server uptime is less that 540 seconds and send notification.
$ crontab -l|grep reboot
##### monitor node reboot #####
*/5 * * * * /home/oracle/scripts/last_reboot.sh 540 > /tmp/last_reboot.cron 2>&1

A more efficient implementation is to run a cronjob automatically after the server restart.

Here is high level example:

### When server is restarted, host_restart_alert.sh will be executed
[root@oracle-12201-vagrant ~]# crontab -l
@reboot su oracle -c '/home/oracle/host_restart_alert.sh' > /tmp/host_restart_alert.out 2>&1

### Here is host_restart_alert.sh
[oracle@oracle-12201-vagrant ~]$ cat host_restart_alert.sh
#!/bin/bash -x
# Script ie being called from root crontab
# uptime reports minutely and need to sleep for at least 60s after host restart
sleep 63
EMAILMESSAGE="$(hostname) was restarted `uptime -p| awk -F'up' '{print $2}'` ago at `uptime -s`"
echo $EMAILMESSAGE > /tmp/restart_$HOSTNAME.log
exit

### Comment from colleague:
### From a bash syntax perspective, it’s not wrong. It’s not great style (don’t use backticks)
printf -v EMAILMESSAGE '%s was restarted %s ago at %s' \
"$(hostname)" \
"$(uptime -p| awk -F'up' '{print $2}')" \
"$(uptime -s)"
echo $EMAILMESSAGE > /tmp/restart_$HOSTNAME.log

### Deconstructing uptime commands:
[oracle@oracle-12201-vagrant ~]$ uptime -p
up 17 hours, 28 minutes

[oracle@oracle-12201-vagrant ~]$ uptime -s
2021-02-15 18:00:51

### Deconstructing message sent:
[oracle@oracle-12201-vagrant ~]$ echo "$HOSTNAME was restarted `uptime -p| awk -F'up' '{print $2}'` ago at `uptime -s`"
oracle-12201-vagrant was restarted  17 hours, 28 minutes ago at 2021-02-15 18:00:51

### Demo:
[root@oracle-12201-vagrant ~]# date
 Tue Feb 16 14:51:18 -05 2021

[root@oracle-12201-vagrant ~]# uptime
  14:51:22 up 1 min,  1 user,  load average: 0.58, 0.23, 0.08

[root@oracle-12201-vagrant ~]# ls -l /tmp/restart
 -rw-r--r--. 1 root   root     271 Feb 16 14:51 /tmp/host_restart_alert.out
 -rw-r--r--. 1 oracle oinstall  71 Feb 16 14:51 /tmp/restart_oracle-12201-vagrant.log

[root@oracle-12201-vagrant ~]# cat /tmp/host_restart_alert.out
 sleep 63
 ++ hostname
 ++ uptime -p
 ++ awk -Fup '{print $2}'
 ++ uptime -s
 printf -v EMAILMESSAGE '%s was restarted %s ago at %s' oracle-12201-vagrant ' 1 minute' '2021-02-16 14:50:02'
 echo oracle-12201-vagrant was restarted 1 minute ago at 2021-02-16 14:50:02
 exit 

[root@oracle-12201-vagrant ~]# cat /tmp/restart_oracle-12201-vagrant.log
 oracle-12201-vagrant was restarted 1 minute ago at 2021-02-16 14:50:02
[root@oracle-12201-vagrant ~]#

Scripts were tested on Oracle Linux Server release 7.8  and 7.9.

Using sed To Search And Replace

The goal is to replace me@gmail.com with dba@gmail.com for all shell scripts.

Fortunately, all shell scripts are located from one directory; otherwise, will need to find all locations.

Check crontab to find possible directory location for shell scripts.

[vagrant@oracle-12201-vagrant ~]$ crontab -l
5 4 * * * /home/vagrant/scripts/test.sh something > /tmp/test.out 2>&1
[vagrant@oracle-12201-vagrant ~]$

[vagrant@oracle-12201-vagrant ~]$ crontab -l|grep -v '#'|grep sh|awk '{print $6}'|sort -u
/home/vagrant/scripts/test.sh
[vagrant@oracle-12201-vagrant ~]$

Check directory for shell scripts.

[vagrant@oracle-12201-vagrant scripts]$ ls -l
total 12
-rwxrwxr-x. 1 vagrant vagrant  25 Feb  4 21:15 dt.sh
-rwxrwxr-x. 1 vagrant vagrant  20 Feb  4 21:14 test.sh
[vagrant@oracle-12201-vagrant scripts]$

Check shell scripts containing emails to modify.

[vagrant@oracle-12201-vagrant scripts]$ grep 'me@gmail.com' *.sh|grep sh|awk -F':' '{print $1}'|sort -u|grep -v edit_email.sh
dt.sh
test.sh
[vagrant@oracle-12201-vagrant scripts]$

Create edit_email.sh to modify email.

[vagrant@oracle-12201-vagrant scripts]$ cat edit_email.sh
for infile in $(grep 'me@gmail.com' *.sh|grep sh|awk -F':' '{print $1}'|sort -u|grep -v `basename $0`)
do
  echo $infile
  sed 's/\bme@gmail.com\b/dba@gmail.com/g' $infile > tmp.$$
  mv tmp.$$ $infile
  chmod 755 $infile
  grep 'gmail.com' $infile
done
[vagrant@oracle-12201-vagrant scripts]$

Run edit_email.sh and verify results.

[vagrant@oracle-12201-vagrant scripts]$ ./edit_email.sh
dt.sh
echo dba@gmail.com
test.sh
export PAGER_EMAIL="dba@gmail.com"
[vagrant@oracle-12201-vagrant scripts]$

[vagrant@oracle-12201-vagrant scripts]$ grep 'me@gmail.com' *.sh|grep sh|awk -F':' '{print $1}'|sort -u|grep -v edit_email.sh

Here is an improvement for the code thanks to Jared Still

Filter basename before sort.

Use grep -il

[vagrant@oracle-12201-vagrant scripts]$ cat e.sh
for infile in $(grep 'me@gmail.com' *.sh|grep sh|awk -F':' '{print $1}'|grep -v basename $0|sort -u)
do
echo $infile
done
for infile in $(grep -il 'me@gmail.com' *.sh 2>/dev/null | grep -v $(basename $0) | sort -u )
do
echo $infile
done