From aa94c8503c5727da8c30712dd55005e4c43c7c89 Mon Sep 17 00:00:00 2001
From: Simon Michnowicz <simon.michnowicz@monash.edu>
Date: Wed, 13 Mar 2019 17:52:52 +1100
Subject: [PATCH] first checkin
Former-commit-id: 14ed73b6d33621195f0cde923ec60751284c0529
---
roles/slurm_sql_bk/README.md | 37 ++++++++++++++++++++++++++++++++++++
1 file changed, 37 insertions(+)
create mode 100644 roles/slurm_sql_bk/README.md
diff --git a/roles/slurm_sql_bk/README.md b/roles/slurm_sql_bk/README.md
new file mode 100644
index 00000000..97491d89
--- /dev/null
+++ b/roles/slurm_sql_bk/README.md
@@ -0,0 +1,37 @@
+This role sets up a cronjob on one Management machine to perform a mysql dump.(via a remote mysql dump to SQL machine)
+This file is then gzipped and a 'scp' done to transfer it to the destination machine for storage.(SQL_BK_DEST_HOST)
+
+As this script is meant to be portable across clusters, some variables need to defined when called.
+
+Architecture:
+ As SQL machine may not be able to access SQL_BK_DEST_HOST, a management machine must do a dump. Then this is scp'd to SQL_BK_DEST_HOST
+ MySQL password is stored in mysql.conf in the BIN dir of Management Machine
+ Only one Management should do the dump for performance reasons
+ A cron job on SQL_BK_DEST_HOST deletes any backups > 7 days
+
+
+EXAMPLE USAGE:
+
+ansible-playbook -i static --limit=monarch-mgmt1 slurm_bk.yml
+
+where slurm_bk.yml contains
+
+- hosts: 'ManagementNodes'
+ gather_facts: True
+ roles:
+ - { role: slurm_sql_bk, create_key: "True", SQL_BK_DEST_HOST: "118.138.234.186", SQL_BK_DEST_DIR: "/mnt/backup/monarch", SQL_IP: "172.16.226.88" , sqlUsername: "slurmdb", sqlPasswd: "{{ slurmdb_passwd }}" }
+
+Meaning:
+create_key: If defined to True then ssh-keygen is called on local machine and then:
+ private key copied to ~/.ssh on management node
+ public key inserted into authroized_keys on SQL_BK_DEST_HOST
+SQL_BK_DEST_HOST: IP number of Machine where we store mysql dumps
+SQL_BK_DEST_DIR: Directory on backup machine for mysql dumps. Note this is owned by ec2-user and is manually created. It should be cluster-specific, i.e. different for each cluster
+SQL_IP: IP number of slurm mysql machine
+sqlUsername,sqlPasswd Mysql username/password, same as in slurmdbd.conf
+
+Other VARIABLES:
+defined in default/main.yml
+BIN_DIR: "/root/bin" #where the backup shell script and mysql.conf exists on Management Node
+SQL_BK_DATA_DIR: "/tmp" # where the inital dump exists on the management node. The file is deleted after a scp so it should not fill up disk
+SQL_USER: "ec2-user" The user account on SQL_BK_DEST_HOST. Normally ec2-user but could be something else for security reasons
--
GitLab