pacemaker | Heartbleed client exploit | Hacking library

Here is what I did to simulate what Ansible does (Based on remarks by @Zeitounator): ssh @ '/bin/bash -c "echo $PATH"'

I get my default PATH as explained in the manual page of bash. In my system sh links to bash.

I see that /etc/profile does the path manipulation that I need. However, it seems that because of the option -c, the bash is not started as login shell and therefore etc/profile is not sourced.

I end up doing the job manually:

I have found a Split-Brain that did not appear in the status of pcs.

The issue was that the agent couldn't see the Oracle listener.

The solution was to:

• Create tnsnames.ora
• Write configurtion
• Reload services

After this the pcs didn't have any errors.

• As mentioned in the question that pacemaker, upon putting node04 on unstandby, pacemaker was demoting node03 and tried to make node04 as the master. It would fail in this task and then later make node03 as the standalone master.

• Since I was doubting that it was picking some old configuration from cib or pengine folder, I even destroyed the cluster on both nodes, removed pacemaker, pcs and corosync and re-installed all of them.

• Even after all of that, the problem still persisted. Then I doubted that maybe the folder permission of the /var/lib/pgsql/ folder on node04 might not be right and started exploring it.

• Only then did I realize that there was an old PGSQL.lock.bak file, which was dated June 11, which means it was older than the current PGSQL.lock file in node03, and thus pacemaker tried to promote node04 and would fail. Pacemaker would not show this as an error in any logs. There is no information about it even on the crm_mon output. Once I removed this file, it worked like a charm.

TLDR;

• Check if there are any PGSQL.lock.bak or any other unnecessary files in the /var/lib/pgsql/tmp folder and remove them before starting pacemaker again.

If you're doing Active/Passive (Active/Standby) clustering, there is no difference between a MongoDB on DRBD vs. MongoDB on any other block device.

If you had multiple active MongoDB's accessing a dual-primary (Active/Active) DRBD device, that's where the potential for corruption would come in.

There seems to be something broken:

• in your regular expression
• in the complexity you are bringing in your loop

With regex_findall you can get a list of all the matches, so you don't need to have multiple regex_findall.

The resulting list would actually be a list of list contains all the matches, then for each matches line, the fragments that you are capturing with the capturing groups of your regular expression.

So, given:

OK, I solved it, changing the resource in pacemaker.

There are a number of possible causes for a high availability queue manager to not run on the node you may expect it to, a common cause is failed resource actions. If you run the crm status command, you may see a "Failed Resource Actions" section which may detail a failed resource action preventing the queue manager from running on its preferred node.

The rest of my answer assumes that you did have a failed resource action (i.e. you see a "Failed Resource Actions" section).

Reading the text surrounding it sometimes gives you a hint that you have an underlying issue that you need to fix. Sometimes you can find more clues about underlying issues from the syslog or dmesg at the time of the failed action. If the failed resource action has an "exitreason", try searching for parts of the text in the syslog and dmesg.

Once you have resolved any outstanding issues (if there were any), clear the failed resource action(s) by running crm resource cleanup RESOURCE, replacing "RESOURCE" with the name of the resource that failed (e.g. 'p_fs_haqm1' or 'haqm1'. N.b. the failed resource action's name will be prefixed by the name of the resource). There may be multiple failed resources, so you will need to issue the command for each of them. Note, if the underlying issues weren't fixed, then the action may fail again, which will be seen by reissuing crm status.

For more information visit https://www.ibm.com/support/knowledgecenter/SSFKSJ_latest/com.ibm.mq.tro.doc/q133450_.htm (remember to "Change version or product" to your MQ version) where you will find a section titled "Pacemaker scenario 2: An RDQM HA queue manager is not running where it should be", which goes into a lot more detail than my answer.

OWN SOLUTION:

Of course, the above-mentioned agglomeration of one SUM(...) after another does not work as this builds up an array of SQL query result sets in rows which do display the associated MIN, MAX and AVG duration for the type (not the sum!) of surgery performed but cannot be displayed without further ado using PHP.

The resulting SQL query code is like this: