A painful move brick summary--online Management Server Migration

Why this migration, mainly because of the years before the expansion of Nagios to do a lot of research and testing, want to apply to the production environment, but the production environment Nagios Server is a centralized Management server, run a lot of open-source software, and most of the predecessors installed deployment, the structure has been cured, Too many pits have been unable to expand; second, the Management Server operating system version for Centos5.4, to be honest now many software installed on the 6x system is more convenient, the default environment basically can meet a variety of open source software operation, and offline testing is 6.5 of the system tested. Finally, because the management server is too old, afraid of which day down, although the configuration files are backed up every day, but the software installed together or more headaches. So finally, we decided to migrate all the services on the Management Server to the virtual machine, so that the backup of the virtual machine after the software installation is complete can resolve the server down recovery problem.

Why this migration is painful, because I spent one weeks on the new virtual machine installed a variety of programs, the migration of various configurations, scripts to solve the installation process of various problems (although many programs have been installed in the local test, but the line still encountered many problems previously not encountered), At the same time also to modify all the relevant configuration of the server, every day to get the faint day surface, headache blurred, mood irritable.

First of all, the migration process has encountered some of the more headache problems.

1.nagios uses Nrpe to monitor each server, but Nrpe has a more perverted configuration allowed_hosts, that is, which IP is allowed to access it, before all the servers are configured with the old Management Server IP, And I studied for a long time found that it cannot configure the IP segment. OK, the new Management Server is another IP, then the dead, all the server Nrpe this parameter to add a new IP, the line has about 30 physical machines, n virtual machines, all come over two hours without.

2. I have a slight obsessive-compulsive disorder, the appearance of the interface is relatively high requirements, nagios4x version of the interface is very good, so the new server on the line installed on the nagios4x version, As a result, both the Pnp4nagios and the ndoutils used are not working (the NAGIOS3X version is used for offline testing). Then to solve these two problems wasted a day of time. ndoutils This most pit father, directly from the note copy of the configuration, resulting in ndoutils has been unable to run normally, and then all kinds of Baidu, Google, look at official documents, finally copied from someone else's document under the same configuration miraculously normal. Later found that the software is defective notes, many records of the command directly copied out of the normal use.

3. Operations Management Platform (we wrote the management platform), there are many migration scripts, migrated to the new server after the test, the result because of the various non-standard written before (my own development level that is called a slag, can run can), change Ah, test ah, and then change, the head is big.

4.jenkins. The previous deployment used the root user, after the migration in order to standardize the use of ordinary users, the results are conceivable, but also a variety of changes, various tests.

5. There are also two services on the management platform: NTPD and NFS, all right, after the migration, it's like Nrpe, all the servers are all over again.

6. Debugging stage: After a few days in the headache brain, all programs, configuration has been migrated, start testing, found all kinds of lost Ah, here forgot to migrate a script, there forgot to migrate a configuration, here forgot to install a RPM package, rush

In view of these problems, summed up a few points:

1. Before the migration of the preparation is too rough, just a simple investigation of the next old server running programs, no record documents, no careful collation, no expansion of thinking, many and the main program related to the small things are in the debugging when found missing just remembered.

2. There is no volume management tool, as long as the move to the public configuration, so good all the server to come over, too hurtful. Automated operations management Tools Learn a few things, and in contrast, prepare to deploy one in the production environment.

3. Specification, before the pit will not say, all the operation to be standardized, so that the migration will not be so painful.

This article is from the "No Technology Madness" blog, please be sure to keep this source http://s8576.blog.51cto.com/9510968/1746409

A painful move brick summary--online Management Server Migration