topic Basic setup (scalable + Serious) in Alfresco Archive

Basic setup (scalable + Serious)

t16 — Tue, 13 Jan 2015 22:18:52 GMT

Is this scenario remotely possible?Two Alfresco web front ends, 2 Solr boxes and 2 Transform servers with load balancing between them?Does the alfresco web front end install separately to the main Alfresco instance, or is the web front end and the instance one and the same, and you need to setup a p

Re: Basic setup (scalable + Serious)

mitpatoliya — Wed, 14 Jan 2015 06:35:29 GMT

Most of the possible combination which you have mentioned above are possible with Alfresco Enterprise.
You can have seperate Alfresco Share instance(WebFront) , Alfresco Repository and seperate Solr server. All of these can run on different instances.
So "one Alfresco box behind the two web fronts" Possible.

Now as far as clustering is concerned you need to keep in mind three things to optimize the perfomance
Indexes, Database and FileSystem.

Different types of setups have different pro and cons, few area of treadeoffs are as follow.
Syncup delay, Downtime (High avilablity),Hardware Requirements, Scalability etc..

So, I guss this will give you good starting point.

Re: Basic setup (scalable + Serious)

mrogers — Wed, 14 Jan 2015 11:27:00 GMT

Answers yes you can do "serious scaling" with alfresco.   If you are really a "total noob" should probably get alfresco consulting or an alfresco partner to advise on your setup.

The choice of what O.S. to run tends to be decided by whatever your admins are most comfortable with.   But I think the preferred platform is NOT windows.

You can also "mix and match" your O.S. to suit for example you can run your MS database on windows. And alfresco on linux.   Its not a problem.   You could even do it in separate VMs on the same box.   Or mix and match cloud computing.

Re: Basic setup (scalable + Serious)

t16 — Fri, 16 Jan 2015 16:25:51 GMT

Hi Guys,

Thanks for the feed back, but its a project for myself as much as anythign else.

Can someone tell me a rough outline to have multiple share servers, a single content server and 2 x SOLR boxes?

I mean, when I install vanilla Alfresco, it installs with Share… How do I install an alfresco instance, with two share front ends separetely?

As it stands I dont know how to install just share on its own, and set it up to communicate to another alfresco server.

The SOLR boxes seem simple enough, one SOLR box dedicated to each share (web) front end?

The two web front ends syncing via Hazlecast?

Many thanks!!

Re: Basic setup (scalable + Serious)

mrogers — Fri, 16 Jan 2015 21:12:00 GMT

Rough outline is load balancer in the front.

Two instances of share.

One repository

And a load balancer in front of your solr nodes. 2 solr nodes is probably overkill.

Installation instructions are in docs.alfresco.com.

Re: Basic setup (scalable + Serious)

t16 — Sat, 17 Jan 2015 03:39:00 GMT

Wow straight from the horses mouth, I am honoured!

I was under the impression that the SOLR indexes had to be tied to each Share/Web front end?

But it seems you are saying that I can just point my two share front ends, at a load balancer with 2 SOLR servers behind it, which will give failover etc for SOLR also?

I assume both SOLR boxes speak to the single repository separately, or do they sync with each other somehow?

Thank you so much for the reply, its really appreciated. I'm starting to really love this stuff, never want to be a developer as I hate coding, but it seems there is so much stuff you can customise with the extensions folders etc that real hardcore coding of the java internals seems rare for us virtualisation guys!

Forgot to mention, we have around 1200 users, not a lot I know, but things can get quite intensive by the nature of their use. So 2x Share + 1x DB + 1x Alfresco Repository + 2x SOLR + 2x Transform Server should give us the scale up performance for the future, AND provide some redundancy for us, in case we need to drop a VM for maintenance etc.

Do you think I am crazy for the above scenario, or do you think this is sensible and forward thinking?! You can be honest, I only started with this stuff a week ago and I'm hooked.

Re: Basic setup (scalable + Serious)

mrogers — Mon, 19 Jan 2015 12:33:26 GMT

Welcome.   No you are not crazy but I'd start with a small working system and then trick it out afterwards.

I don't think there's a direct link between solr and the front end.    (Its possible the recent 5.0 stuff has added one but AFIK that's not the case.)     So with one repository node you need the LB to spread out the load to multiple SOLRs.

When you start clustering the repo nodes (Enterprise only) you can play with other patterns. For example the last time I looked at the alfresco cloud it was 4 repo nodes serving the front end.    With two solr nodes behind and each solr node had a dedicated private repo node for tracking.

Re: Basic setup (scalable + Serious)

t16 — Mon, 19 Jan 2015 14:16:00 GMT

HI guys I have another question im really sorry…

If I have Hazelcast enabled between two share front ends, is that enough to provide the Tomcat session replication between the two instances, or do I need to have hazelcast running, AND tomcat session replication configured also, or does Hazelcast do the whole lot?

THanks!!

Re: Basic setup (scalable + Serious)

mrogers — Mon, 19 Jan 2015 15:19:23 GMT

Follow the docs for clustering of http sessions.

You just need to configure share to use hazelcast and off it goes!

Re: Basic setup (scalable + Serious)

t16 — Mon, 19 Jan 2015 17:42:00 GMT

!!

IS NFS an acceptable way of sharing a content store, or is something more serious like GlusterFS recommended?

If I have an existing local content store setup, and I want to point it at a shared store, do I just copy the files over from the local one to the shared one, or is there more to it?

Thank you so much for all your help. Im starting to get seriously into Alfresco, I really like and enjoy it, apart from the fact Im not anywhere near to the stage of coding up custom things, more just focused on the architecture and suchlike!

Re: Basic setup (scalable + Serious)

t16 — Wed, 28 Jan 2015 13:18:07 GMT

HI MRogers,

I have a small but significant question:-

I am proxying sucessfully with a load balancer, but I notice that when I edit online, of course I have a direct connection to the alfresco host itself, and this means that its bypassing the proxy and the compression I am applying to the data.

Can I just point the alfresco VTI server to the load balancer, and then that will forward the requests back to the alfresco box, or will this introduce a loop, thus doubling the network traffic??

With an HTTPS front end, I did try this (SSL termination at the balancer) but it always failed to connect through the LB. Even if HTTPS was set in the protocol on the alfresco box. For some reason it was always coming in as 443 when this is enabled, and was not applying the :7070 port setting.

Anyway I can fix that, the bottom line is, can I point the VTI server setting BACK to the Balancer in front of the Alfresco cluster, or is this a very bad idea? We need compression because of some slow WAN links!

MAny thanks for all your help.