ATOL: Heartbeat

29.6.2009 11:38 | Přečteno: 1663× | Linux

Author: Pavel Piskač

Heartbeat is part of the Linux-HA project, where HA means High Availability. Heartbeat is a daemon that provides cluster infrastructure (communication and membership) services to its clients. This allows clients to know about the presence (or disappearance!) of peer processes on other machines and to easily exchange messages with them.

In order to be useful to users, the Heartbeat daemon needs to be combined with a cluster resource manager (CRM) which has the task of starting and stopping the services (IP addresses, web servers, etc) that cluster will make highly available.

Heartbeat comes with a primitive resource manager, however it is only capable of managing 2 nodes and does not detect resource-level failures.

A new resource manager which addressed these limitations and more was written for Heartbeat 2.0.0. However in 2007 the new resource manager was spun-off to become the Pacemaker project in order to better support additional cluster stacks (such as OpenAIS) and is no longer associated with the Linux-HA project.

The current stable series of Heartbeat is 2.99.x and can be obtained for many platforms. Because of high portability, users can use this service on GNU/Linux, Solaris, OpenBSD, even on MacOS/X from time to time.

What Heartbeat provides

To be able to use Heartbeat on very wide systems, there is no fixed maximum number of nodes. Very small a simple clusters can be of course build using Heartbeat too. Each cluster has to have some resource monitoring which can automatically restart move resources to another node on failure.

Another cluster automatic healing function is fencing mechanism. Fencing is the component of cluster project that cuts off access to a resource (hard disk, etc.) from a node in your cluster if it loses contact with the rest of the nodes in the cluster.

The most effective way to do this is commonly known as STONITH, which is an acronym that stands for "Shoot The Other Node In The Head." In other words, it forces the system to power off or reboot. That might seem harsh to the uninitiated, but really it's a good thing. If a node that is not cooperating with the rest of the cluster can seriously damage the data unless it's forced off. So by fencing an errant node, we're actually protecting the data.

Fencing is often accomplished with a network power switch, which is a power switch that can be controlled through the network. This is known as power fencing.

Fencing can also be accomplished by cutting off access to the resource, such as using SCSI reservations. This is known as fabric fencing.

For resource managing, there is sophisticated policy-based resource management, resource inter-dependencies and constraints. If your cluster has regular power peaks, you should use time-based rules allow for different policies depending on time, which add extra nodes in case of they are needed or it is able to decrease number of nodes in case the cluster does not need them.

Each cluster can provide some special service or set of services. Developers prepared several resource scripts (for Apache, DB2, Oracle, PostgreSQL etc.) to make using these services easier.

Some users prefer graphical user interface instead of editing text files. These user will welcome GUI for configuring, controlling and monitoring resources and nodes.

How to configure your cluster

First of all, you have to install Heartbeat. If you don't have some rare distribution, your repository should contain Heartbeat package. In other case, compiling source codes takes place. Examples how to install Heartbeat on Debian and CentOS.

aptitude install heartbeat-2

yum install heartbeat

Compiling source codes can be said as rare and it is described in official documentation. There are created some directories during installation. The most important of them are

/etc/ha.d

and

/usr/share/doc/heartbeat-???version???

. In the first of them you will find almost nothing but configuration files should be there. Examples of these files are places in the second directory.

Following configuration files are needed to make the cluster to work. We will start with the most important of them,

ha.cf

, this is the main configuration file where are set services and cluster behavior. Second configuration file is

haresources

, where are saved nodes with their resources. These resources move from machine to machine as nodes go down and come up in the cluster. The last of them is

authkeys

where is saved authentication information. For those who don't read manuals. There is a password which must be the same on all nodes in the cluster. Each file contains rich guide with examples with some regular configuration.

The will not be mentioned any set of practical examples because they can be found in each configuration file. Moreover, there are lots of cluster types with many services and to be able to mention at least small part of them, it would be little book.

Pacemaker

Pacemaker makes use of your cluster infrastructure (either OpenAIS or Heartbeat) to stop, start and monitor the health of the services (aka. resources) you want the cluster to provide.

It can do this for clusters of practically any size and comes with a powerful dependency model that allows the administrator to accurately express the relationships (both ordering and location) between the cluster resources.

Its is also worth reiterating that Pacemaker is NOT a Fork of Heartbeat, as it seems to be a common misconception.

Pacemaker uses configuration file which is saved in

/var/lib/heartbeat/crm/cib.xml

. This file can be created using following command

/usr/lib/heartbeat/haresources2cib.py /etc/ha.d/haresources

You should never edit this file by hand! This file is changed according to actual cluster state and you should only create it from

haresources

using

haresources2cib.py

. That's all you have to do with this file.

Conclusion

There are some advantages and disadvantages, I will try to describe the most important of them.

To able to ensure, that the cluster is working properly, you can use init script to start, stop or restart the service. It will help you with finding out, how is the behavior of the whole cluster and its services if some of the nodes die.

Human readable configuration files are very important for many administrators. There are many examples and documentation. But there is bad news for those who don't like xml configuration files, developers are planning to use them in the next version of Heartbeat.

Heartbeat is wide-spread cluster application and there are many guides, user's experience and documentation on the Internet. Both newbie and expert will find solution of his problem.

If you will use some wide-spread application (Apache, DB2, Oracle, PostgreSQL etc.), you can use prepared scripts for their management in cluster. If there is no official script, community certainly created something, what you can use.

Not to be so optimistic, there are also some disadvantages. Heartbeat is not suitable for people who don't read manuals. There are many options whose function is not clear and the cluster will not work properly without knowledge of the right configuration. Because of some functions which were mentioned above, the first configuration can be a little bit difficult for newbie.

The biggest disadvantage for me is changing configuration files from classic text files to xml. In my opinion it is not good choice but developers have another reason and they hopefully know what they do.

Links

http://linux-ha.org/

http://clusterlabs.org/wiki/Main_Page

http://en.wikipedia.org/wiki/Heartbeat_(program)

http://sources.redhat.com/cluster/wiki/FAQ/Fencing

       

Anketa

Komentáře

[Michal Fecko]

你為什麼把它寫在英語？

[AsciiWolf]

To musíš pod každým jeho článkem takhle trollovat?

Proč by to nemělo být v angličtině? Jsme na technickém portálu a proto předpokládám (a nejen já), že většina lidí zde bude schopna přelouskat pár jednoduchých anglických vět.

Možná autor tohoto blogpostu nepíše naprosto správně/spisovně anglicky, nicméně jeho blogposty jsou kvalitně a jednoduše napsány, takže to imho nemá problém přečíst i člověk se základy (technické) angličtiny.

Pokud ti to opravdu dělá problém, tak se k tomu nevyjadřuj...

[ATOL @ FI]

Je to blog predmetu, takže od jedného autora je obvykle len jeden článok :)

[xxxs]

mozno by si mal pred kazdym zapiskom nahodit male vysvetlenie v slovencine.

[Marek 'marx' Grác]

To by asi dávalo zmysel a súčasne by bolo jasnejšie z akého predmetu to je.

[xxxs]

ozaj, aka bola kofola?

[pools]

中国是一个很好的语言

[AsciiWolf]

A navíc. Proč to píše anglicky, napsal Zde (do svého prvního blogpostu).

[mimozemšťan]

Já příští blog napíšu mimozemsky.

[=^..^= AmigaPower®]

...a já nizozemsky.

[Michal Fecko]

A ja v činsštine

[=^..^= AmigaPower®]

...a v jakém dialektu?

[.]

您好，我很高興，您知道如何中國

[m-a]

Aspon po slovensky keby to bolo spravne.

[xindl]

a já pičo v příšerné češtině ať se všichni poserete!

[.]

No, kluci. Čumte jak strejda Google za 0,876543s přeložil do češtiny.

[=^..^= AmigaPower®]

Já obrovský pinďour mít v ruka ráno dnes svůj!

[stativ]

Mr Happy, I have a huge tangan di pagi hari ini!

[.]

ty máš kocourův pinďour?

[=^..^= AmigaPower®]

Ty ne?!

[Michal Fecko]

However in 2007 the new resource manager was spun-off to become the Pacemaker project in order to better support additional cluster stacks (such as OpenAIS) and is no longer associated with the Linux-HA project.

preklad a la Google

Nicméně v roce 2007 nový zdroj manažera byl opředený-off, kterým se stal kardiostimulátorů projekt s cílem lépe podporovat další cluster komíny (např. OpenAIS) a již není spojena s Linux-HA projektu.

[Nikola Ciprich]

If I can recommend, rather look at openais/corosync, as this is the stack of choice for future pacemaker development, not the heartbeat.

[hikikomori82]

V ankete mi chyba moznost: I don't give a shit

[Adam Rohac, Roznava]

English sucks!