Hi Leonel. That would be a harder task. I guess you should be able to infer something about the newness of your data by checking the binary log files. The highest number or newest date would be my guess, but I've never had that challenge. I hope this helps. Thank you for watching my videos. Best regards Daniel
thank you for this! question: why did you use as wsrep_sst_method = rsync instead of mysqldump or xtradb ?i saw in a other video (from someone else.. which was older) that of course everyone has pro and contra but xtradb would have the least lock time on the tables ? also: would you mind combine this video with another where you use haproxy or other loadbalancer and explain why you used them ? what you've done is good for redundancy but if someone wants to make a cluster with every node to be active/master ?
Hi Fritz This video was more of an introduction with all the information I grained when I researched it as an option for our migration. Didn't do a deep dive and never saw the option extradb. Not mentioned in this documentation either. galeracluster.com/library/documentation/mysql-wsrep-options.html#wsrep-sst-method We haven't given up on Vitess yet but there might be a reason for us to consider Galera again and then I must figure out the right HA setup options. Thank you for watching my videos. Best regards Daniel
Hi Daniel! Thanks for the video. I need to know how many tps maximum I can expect with this configuration. I saw some videos that talk 1million qps. I think that isn't possible in production environments.
Hi Andres. That is pretty impossible to answer, you need to test it yourself. The parameters to handle is speed of ram, drive speeds and general CPU utilization of the machine. If you have a really fast machine, ram and disk you can achieve fast queries and even improve speed if you can do them in parallel. I don't have any tooling to benchmark databases but I might find one in the future. Thank you for watching my videos. Best regards Daniel
Hi Maybe you have an older version of mariadb or perhaps you've installed mysql. Hard to know what is wrong without more information. Thank you for watching my videos. Best regards Daniel
Thanks for this video. I have recently had an outage on my DB server, so now I'm running two MariaDB servers on primary and replicate configuration. My question is: I have lots of services that access the primary DB server, if that one goes down, how should I setup things for the services to automatically switch to the replicate server? At the moment I defined a domain with the local IP of the current primary, so I can change the register of that DNS and every service would resolve the new IP.
Hi Thank you for watching my videos. At work we use heartbeat a lot, there is also more complicated solutions that do similar things but you could specify an static IP for your server that the clients connect to and then if one server goes down the other one will realize and heartbeat will assign the IP. This video covers heartbeat in general: th-cam.com/video/Nq8K9SPAODc/w-d-xo.html I hope this helps. Best regards Daniel
Hi. That is not something I've looked into. I guess that could be useful in some cases. I would google for an answer. I hope you find your answer. Thank you for watching my videos. Best regards Daniel
Thnx for the video, i had one question though, After setting the global var, wsrep_provider_options='pc.bootstrap=true', would it be wise to set is to false after the cluster is up and running again, to prevent that this node will allways be the bootstrap-one?
Hi Sangre Yes, that is a really good point. Especially if you need to upgrade / restart a node. So these variables needs to be maintained. Thank you for watching my videos. Best regards Daniel
Hi Shrine. Yes, if the remaining nodes are in a quorum you should be able to start the third node and it should pick up where it left off. Might take a while before it has catches up. I hope this helps, thank you for watching my videos. Best regards Daniel
@@DanielPersson That's correct but I'm mentioning situation when 1 of 3 node fails (or is down) and 2 left are running. Is it safe to restart any of those 2 remaining nodes?
Hi Shrine. Ohh, I see how I misunderstood your question. Well in that case I would not restart it without handling it as a "one node left case". It's a complicated case because if you don't want to loose data or connectivity I would much rather get the third node up and running and then restart the first two. Best regards Daniel
Hi Well I did all the steps required to install it on a Debian machine. All packages are included in a normal maria-db installation. Load balancing is a different topic but the servers will be in quorum so you could connect to any of them. In my installation I would put a haproxy in front, I have a video about that but it's not required for the service to be useful. Thank you for watching my videos. Best regards Daniel
This video is absolutely gold. Thank you, it helped me a lot.
Hi Daniel,thank you very much for the time you have dedicated in making this super useful video..cheers from Italy.
This was a very interesting demonstration. Thanks!
thanks for the video, this made some knowledge about Maria DB HA
Thanks
Tks for your training, but I wonder how can I detect which server contain newest database in reality when all servers in cluster died?
Hi Leonel.
That would be a harder task. I guess you should be able to infer something about the newness of your data by checking the binary log files. The highest number or newest date would be my guess, but I've never had that challenge.
I hope this helps. Thank you for watching my videos.
Best regards
Daniel
thank you for this! question: why did you use as wsrep_sst_method = rsync instead of mysqldump or xtradb ?i saw in a other video (from someone else.. which was older) that of course everyone has pro and contra but xtradb would have the least lock time on the tables ?
also: would you mind combine this video with another where you use haproxy or other loadbalancer and explain why you used them ? what you've done is good for redundancy but if someone wants to make a cluster with every node to be active/master ?
Hi Fritz
This video was more of an introduction with all the information I grained when I researched it as an option for our migration. Didn't do a deep dive and never saw the option extradb. Not mentioned in this documentation either.
galeracluster.com/library/documentation/mysql-wsrep-options.html#wsrep-sst-method
We haven't given up on Vitess yet but there might be a reason for us to consider Galera again and then I must figure out the right HA setup options.
Thank you for watching my videos.
Best regards
Daniel
🤝
great video thanks. resources are hard to find on this subject.
We used to say: "Aonde está a galera ?" that means: How about the guys ? Where they are ? Something like this.
❤ Next video: configurate maxscale, haproxy, corosync and pacemaker
Hi
Thank you for watching my videos.
I've added the idea to the list of ideas.
Best regards
Daniel
Hi Daniel! Thanks for the video. I need to know how many tps maximum I can expect with this configuration. I saw some videos that talk 1million qps. I think that isn't possible in production environments.
Hi Andres.
That is pretty impossible to answer, you need to test it yourself. The parameters to handle is speed of ram, drive speeds and general CPU utilization of the machine. If you have a really fast machine, ram and disk you can achieve fast queries and even improve speed if you can do them in parallel. I don't have any tooling to benchmark databases but I might find one in the future.
Thank you for watching my videos.
Best regards
Daniel
Hi Daniel , I'm doing this in ubuntu sysrtem and i have come far but while executing sudo galera_ it says that commant not found
Hi
Maybe you have an older version of mariadb or perhaps you've installed mysql. Hard to know what is wrong without more information.
Thank you for watching my videos.
Best regards
Daniel
Thanks for this video.
I have recently had an outage on my DB server, so now I'm running two MariaDB servers on primary and replicate configuration. My question is:
I have lots of services that access the primary DB server, if that one goes down, how should I setup things for the services to automatically switch to the replicate server?
At the moment I defined a domain with the local IP of the current primary, so I can change the register of that DNS and every service would resolve the new IP.
Hi
Thank you for watching my videos.
At work we use heartbeat a lot, there is also more complicated solutions that do similar things but you could specify an static IP for your server that the clients connect to and then if one server goes down the other one will realize and heartbeat will assign the IP.
This video covers heartbeat in general:
th-cam.com/video/Nq8K9SPAODc/w-d-xo.html
I hope this helps.
Best regards
Daniel
it's possible to exclude certain databases?
Hi.
That is not something I've looked into. I guess that could be useful in some cases. I would google for an answer.
I hope you find your answer. Thank you for watching my videos.
Best regards
Daniel
Thnx for the video, i had one question though, After setting the global var, wsrep_provider_options='pc.bootstrap=true', would it be wise to set is to false after the cluster is up and running again, to prevent that this node will allways be the bootstrap-one?
Hi Sangre
Yes, that is a really good point. Especially if you need to upgrade / restart a node. So these variables needs to be maintained.
Thank you for watching my videos.
Best regards
Daniel
So I in case of 3 nodes, when one node fails I shouldn't restart any of remaining nodes without setting pc.bootstrap=true?
Hi Shrine.
Yes, if the remaining nodes are in a quorum you should be able to start the third node and it should pick up where it left off. Might take a while before it has catches up.
I hope this helps, thank you for watching my videos.
Best regards
Daniel
@@DanielPersson That's correct but I'm mentioning situation when 1 of 3 node fails (or is down) and 2 left are running. Is it safe to restart any of those 2 remaining nodes?
Hi Shrine.
Ohh, I see how I misunderstood your question. Well in that case I would not restart it without handling it as a "one node left case". It's a complicated case because if you don't want to loose data or connectivity I would much rather get the third node up and running and then restart the first two.
Best regards
Daniel
please make a detailed video on this. please it would really mean alot. thank youuu
Hi
What detail in the video are you missing?
Thank you for watching my videos.
Best regards
Daniel
Thanks. But you did not show the procedure of installing gelara. And also load balancers
Hi
Well I did all the steps required to install it on a Debian machine. All packages are included in a normal maria-db installation. Load balancing is a different topic but the servers will be in quorum so you could connect to any of them. In my installation I would put a haproxy in front, I have a video about that but it's not required for the service to be useful.
Thank you for watching my videos.
Best regards
Daniel