Hello,
I thinking to install moodle with high availability on my organization.
I need documentation of how to plan and implement this.
Any link?
Thanks very much.
There's been quite a bit of discussion in the Servers & Performance section. Some of the Moodle "Big Hitters" like Catalyst IT have the most experience.
Remember that you almost always have a single point of failure, but relative simple stuff like hot swap drives, sensible RAID arrangements and a spare server chassis cover a lot of possibilities. At the risk of stating the obvious, excellent backup provision is vital.
Remember that you almost always have a single point of failure, but relative simple stuff like hot swap drives, sensible RAID arrangements and a spare server chassis cover a lot of possibilities. At the risk of stating the obvious, excellent backup provision is vital.
Actually I have some experience of government work and one of the things that was pounded into my brain was redundancy. I'm working towards moving to something like the following for HA. I envision having 2 webservers and 2 database servers (slave-master). The single point of failure would be the box that splits the traffic between the 2 webservers, the internet cards, ISP, etc. If I were really paranoid I might go with dual NIC cards and 2 separate ISPs but these are my initial thoughts. Obviously a good RAID setup on each of the servers. As Howard says - a good backup schedule is paramount. Peace.
I think you have to keep a sensible perspective on it. If you ask the boss, he will say "I need 24/7 100% uptime". The reality is that nobody dies if your Moodle server falls over. You can probably spare the half hour to swap out your machines and load the backups without getting fired. IMHO keeping it as simple as possible has some big benefits. Just my $00.02 as they say.
> As Howard says - a good backup schedule is paramount.
Even more important that a good backup schedule is a tested restoration procedure, to make really sure you can bring your complete site back from those nice backups, in a timely and reliable way
Saludos. Iñaki.
Even more important that a good backup schedule is a tested restoration procedure, to make really sure you can bring your complete site back from those nice backups, in a timely and reliable way
Saludos. Iñaki.
Hello,
Thank you very much for your interest.
We are planing an installation like shown in the anexed file (sorry for this tarzanian english ).
You think this configuration can work ?
Greetings
My initial reaction is that the single MySql server will almost certainly be a bottleneck.
What are you actually trying to acheive here? The only duplicated thing is the actual web servers, plus you have a lot of complication. Complication === possible unreliability.
What are you actually trying to acheive here? The only duplicated thing is the actual web servers, plus you have a lot of complication. Complication === possible unreliability.
Hi
Your setup is very interesting.
Just some clarifications:
- The aim is _high availability_ I unterstood. The two servers web1 and web2 are "in parallel". Do they also work in parallel (thereby increasing the throughput) or does only one server answers the HTTP requests, the second one is waiting until the first one fails?
From the comment "PROXY WEB balanced" I think it is the latter case, just to make sure.
- The moodledata/ on NFS: What is the NFS Server? A fifth machine in addition to the four on the diagram?
As already pointed out by Howard the MySQL server is a single point of failure. What is the reason for seperating it from web1 and 2? I mean, why don't you run the MySQL daemon on those machines?
- Why do you have an NFS _and_ SAN? Aren't they competing technologies? Why do you use NFS for moodledata/ and SAN for MySQL.
Although not discussed here ultimately the performance-issue will pop up. Have a look at the relevent forum "Servers and performance": http://moodle.org/mod/forum/view.php?id=596 and if the discussion digress consider shifting there.
P.S.: Lot of participants in moodle.org come up with their real identity. There must be a reason for that.
Your setup is very interesting.
Just some clarifications:
- The aim is _high availability_ I unterstood. The two servers web1 and web2 are "in parallel". Do they also work in parallel (thereby increasing the throughput) or does only one server answers the HTTP requests, the second one is waiting until the first one fails?
From the comment "PROXY WEB balanced" I think it is the latter case, just to make sure.
- The moodledata/ on NFS: What is the NFS Server? A fifth machine in addition to the four on the diagram?
As already pointed out by Howard the MySQL server is a single point of failure. What is the reason for seperating it from web1 and 2? I mean, why don't you run the MySQL daemon on those machines?
- Why do you have an NFS _and_ SAN? Aren't they competing technologies? Why do you use NFS for moodledata/ and SAN for MySQL.
Although not discussed here ultimately the performance-issue will pop up. Have a look at the relevent forum "Servers and performance": http://moodle.org/mod/forum/view.php?id=596 and if the discussion digress consider shifting there.
P.S.: Lot of participants in moodle.org come up with their real identity. There must be a reason for that.
There is some great documentation out there that covers High Availability and Linux Virtual Server.
A couple of links in case you haven't been able to google up the goods
http://www.linuxvirtualserver.org/whatis.html
The High Availability Linux Project
Wikipedia Overview of Clustering and HA
If you want to throw money into the mix, the guys over at RedHat have a pretty good idea about HA Clustering and have a good 16page PDF whitepaper outlining their methodology.
http://www.redhat.com/whitepapers/rha/RHA_ClusterSuiteWPPDF.pdf
A good launcing point and some interesting reading.
Cheers
Brett
A couple of links in case you haven't been able to google up the goods
http://www.linuxvirtualserver.org/whatis.html
The High Availability Linux Project
Wikipedia Overview of Clustering and HA
If you want to throw money into the mix, the guys over at RedHat have a pretty good idea about HA Clustering and have a good 16page PDF whitepaper outlining their methodology.
http://www.redhat.com/whitepapers/rha/RHA_ClusterSuiteWPPDF.pdf
A good launcing point and some interesting reading.
Cheers
Brett