Introduction to ha high-availability software in linux in China

A Brief Introduction to ha high-availability software in linux in China-general Linux technology-Linux technology and application information. The following is a detailed description. 1. vcs. I think bz has many misunderstandings about vcs. I know that zte/hw has basically used vcs in most telecom projects since, and basically all of them are core applications, there is no doubt about the stability of vcs. The only drawback is that it is expensive. In addition, vcs need suse non-standard kernels, which makes me very depressed.

2, Red Flag ha, someone has the problem here, I said, the earliest Red Flag oem called rsf-1 ha software, this is very rough, and it was quite unstable (a user complained about death in a postal project), and later changed the cluster pro (lite) of oem nec. This function was also available and used a few years ago, at that time, I was very depressed because I only had a for windows client, but now I have a web interface-based client.

3. turboha and turbo ha went well in China 05 years ago. In some non-critical zte projects, due to cost considerations, some turbo ha6.0 products were used, later, major developers left, and 6.5 was launched for a long time. It was not a very mature product.

4. The reason why rhcs/heartbeat put the two ha together is that they share the same characteristics and have extremely fancy functions, but it is not the case when it is actually used. hb's resource. d n years have not been updated. You need to change the resource most of the time. the script in d. rhcs now uses the openais stuff for the underlying cluster membership, and the upper-layer script also becomes the same as heartbeat, using the ocf standard, but what I can't bear is, now, after the stop operation of each script is executed, sleep 10 s is required, which is a very sb design. Gfs is used in some commercial bureaus in China, and ut is used in Hangzhou Netcom iptv. Its functions and performance are also good, but it is also a lot of problems. In addition, novell is said to be pushing heartbeat to fight against rhcs. I am not optimistic about it.

5. fujiitsu & simense: This ha software has just entered China. I have done some tests and evaluations. After all, it is done by Germans. It is very rigorous and standardized, but it is just quite reasonable, this software contains the loading balance function module, one of the few linux server Load balancer solutions I have seen is not based on lvs (ibm was edge server seems to also have load balancing functions, but this is common)

6. Because of the price, there are still a lot of people using roseha at the low end. It is said that the previous version is very bad, but after 2.0, the function is still available and the configuration is still convenient. He also has a product called mirror ha, which can be used as a file-based disk image.

7. skybility ha, a dual-host software launched by novell in China because novell did not have its own ha products. It was developed by a company in Shenzhen and its performance was quite satisfactory, the price is also very advantageous. With the support of suse in the operating system, he won a lot of orders for using suse. However, after novell promotes hb, he does not know what will happen. Skybility ha has been used in some edge applications of zte.

8. I have always admired lifekeeper as a professional software. It is a pleasure for programmers to look at the perl code of resource management. His disk image features, performance, and reliability are also very good (based on drbd or nbd + raid1). Unfortunately, lifekeeper is about to leave the Chinese market. It may be too expensive to accept it.

9. legato aam, I have heard of it, but I have not studied it in detail. legato itself is doing well on unix/windows, but not many legato are used in linux in China.

10. shaolin and shaolin are well-known. The product of a Hong Kong company is said to have a R & D organization in Beijing, but has never seen it in China.

11. polyserve. This can also be used as ha software, but it is not as good as its parallel file system pvfs. A company uses a large number of polyserve in iptv projects, create a san + pvfs solution. If his price is low, it is estimated that there is no market for gfs.

12. mc/sg. I will not comment on it if it has never been used.

13. The ha software of hacmp and ibm has been transplanted to linux, but currently only the for power version is available. If aix is used, you should be familiar with it, so I will not comment on it. The difference between Linux and aix is not great.

14. tsa: This is the ha component in Tivoli. I have never heard of it, but I have never heard of it.

In fact, the linux operating system/server is stable enough. As long as the ha software is not messy, users generally don't have to worry about it. Buying the ha software is just to buy a guarantee and psychological comfort. Of course, the necessary functions should still be available. during testing, network cables, power loss, and process killing will usually be tested and tested, which is actually very unreliable. I have seen a power failure, the entire file system on the storage fails and cannot be recovered, resulting in the loss of several TB data.