options for upgrading to 11.4.5

[brian.schnautz]

Hello,
I would like to upgrade us to the latest community edition of Scalix 11.4.5. We are currently running 11.4.1.
I think there are several ways to do this. I personally think the cleanest way to do it would be to migrate the user's mailboxes (7) to a fresh install of 11.4.5. Plus I'm trying to get my Scalix server off of bare metal into a VM. Right now I back up all the user mailboxes daily within a script with:

Code: Select all

sxmboxexp -u "${user}" -a ${MBOXDIR}/"${user}".mbox

Now since I have all the mailboxes in the scalix mbox dump format, I think I can create all the user mailboxes in the fresh install and then simply pump each user's mailbox into the new mailbox with:

Code: Select all

sxmboximp -a /path/to/file/user.mbox --user "John Doe"

Can I please get anyone's opinion if doing this will turn out successful or if there is a more preferred method. I just know that upgrading can get real messy and I would like to keep my new install as clean as possble.
Thanks,
Brian

[Valerion]

Personally I would just migrate the mailstore over and let the installer handle the upgrade. Then you don't have to redo a lot of the configuration files in the system. I've had very little bad experiences, as long as you're careful.

1) Install the new system with the same hostname and IP, and create a Scalix user with the same UID and GID. Also create a Scalix group (check the live server first)
2) Copy /var/opt/scalix over, preserving permissions and ownership
3) Install Scalix and let it upgrade the mailstore
4) Test carefully

Also, be careful of VM's. For low-load systems it's not an issue, but disk IO is the first bottleneck you'll encouter. If you run Scalix in a VM over something like FC SCSI or iSCSI it'll mostly likely run very well. If you're running the mailstore inside the VM itself, you'll likely get issues with more than a handful of users, though.

[brian.schnautz]

Valerion,
I only have 7 users .. and will only have 10 max.. possibly way in the future as many as 15. So do you think putting everything in the vm would be ok considering that?
Brian

[Valerion]

Utimately it depends on your load. You will have to see if it works. It may very well be fast enough for your purposes. You really have nothing to lose, as you can always migrate back to bare metal.

Personally I would rather keep a machine for this. I tend to use VM's only when I do tests for myself.

[kanderson]

Brian, I can't see how 15 users could possibly overwhelm an ESX server.

I have ESX server installs well into hundreds of users.

Past about 50 or 75 users, disk performance will become an issue, as Valerion said. Use SAS drives, in a RAID 10 configuration wherever possible. This will make a large performance benefit for your server. The more spindles you have, the faster it'll go. RAID 10 with 8 drives is almost double the speed of RAID 10 with 4 drives. So if it's a single server, with all the disk in the box, then stuff in as many drives as possible, and use RAID 10. Also if possible, use 15K SAS. If you're only using a small server, and are using SATA drives, you'll see scalability cut dramatically, but at least go for the 7200RPM drives rather than 5400, and here, RAID 10 is a must. Avoid RAID 5 at all times whenever possible.

Also, Virtual Machines should ALWAYS be built as single processor units. MultiProcessors will RAPIDLY degrade the speed of the server overall, and really undercut the benefit of VMware. You'll see better performance with 1 virtual CPU in almost all cases.

If you can, one alternative that you might be interested in (and Valerion hints at it) is using iSCSI storage. To do this in a cost effective way, have a look at openfiler. Note that this is NOT the equal of a hardware SAN. But it'll give you shared storage, which will be really important later on as your company grows, and (strangely) I find openfiler to be faster than local disk as well as much more flexible. You should be able to attach a LUN directly, but give that some thought. I personally doubt that 15 users will notice the speed, and if you directly attach a LUN, you'll lose the ability to use some of the vMotion tools.

Kev.

[brian.schnautz]

I'm just simply using vmware server 2.0.. nothing fancy here.. just something to take a snapshot before I do an updates and such.. that way I don't completely hose my server. I don't have ESX and all the bells and whistles of fully purchased vmware enterprise.

But anyway...Thanks for all you help! I greatly appreciate it!

I'm hoping for speed increase from 11.4.1 to 11.4.5. It literally takes almost a minute to load my 750 company contacts (public folder) through the scalix web interface

[kanderson]

Ok, you have a couple of things here.

1) Vmware server vs esx.
Vmware server is a waste of your time. It'll never scale, and it's not of much use, at least not for running server applications. You can download ESXi for free, and you should do so. It's a VASTLY better product, and it IS suitable for use on a server. I would STRONGLY advise using that as a first step. You'll get SOME of the bells and whistles, and the cost is the same. You'll also avoid the underlying OS overhead that you have with VMware server. Plus, you'll be able to learn about the Enterprise products, so that when the time comes that you do work with them, they'll already be familiar. You CAN upgrade your VMserver images to ESXi. Do the research first though.

2) SWA is slow.
This is almost always a PostgreSQL problem. The easiest way to check if Postgres is working is to sign into your mobile client. http://server.domain.com/m I suspect it'll fail. If it succeeds, then postgres is running and allowing connections. I suspect it won't be though. If that's the case, search the forums here to see if you can fix it yourself. If not, post again with the errors you're seeing in the scalix postgres logs. As a point of reference, SWA loads a new folder for me in about 5 to 10 seconds, and I have a LOT more than 750 entries in some of my folders.

3) This is back to Valerion's disk io comment. Assuming you're on CentOS/RHEL, type "yum install sysstat", to install iostat. Then, monitor disk use while the server is running for a while. You can use iostat -x -d 1 1000. That says give me detailed stats for each device every second for 1000 seconds. This will show if your disk utilization is approaching 100%. Again, adding more disks to the array, or moving to a direct connection will be your best fixes here. ESXi will be faster than VMware Server, as well.

[brian.schnautz]

kanderson,
thank you for all your advice. I don't know if my server will run ESX because of the hardware requirements being so specific. I'll try. Either that, or I might try running Citrix/Xen Server.. They have a free product that runs very well according to friends of mine going on the cheap...and it's not so specific regarding hardware other than needing x86_64 architecture.

I can load https://mydomain.com/m ...but it took about 30 secs or so just to log in. I will follow your advice in your above message and by the way.. i'm not virtual now..but am intending on doing so. Scalix is installed on bare metal currently, but I would like to virtualize it for the sake of taking snapshots and portability.

If I was searching through the forums, what key words would you recommend searching with to resolve my 'slowness' issues. I've tried looking before, but I don't think I was searching for the correct terms..because the results were way too broad. I love scalix, If it wasn't for the speed issue, I think it's a great product!

Thanks again.

[kanderson]

If the mobile client works, then Postgres is likely NOT your issue. Is SWA fast once it's open and being used?

For VMware, the main problem people have with ESXi is the raid controller. If you use iSCSI, (openfiler) then this won't matter at all. If you use a built in RAID controller, post what it is, and we'll see. Most of them have workarounds that let them work fine.

I have never used Xen for one of my servers, though I did support someone using it once. It worked fine, but was a pain to upgrade, as I remember it. Scalix support will be mostly familiar with VMware as well. I believe they use it internally. I don't know anyone using HyperV. You'll need 64bit for ESXi 4 but ESXi3 will run on 32 bit hardware. I'd advise 64 bit anyway though...

Personally, I'd start by looking in the various logs. Check in /var/opt/scalix/??/tomcat/logs and /var/opt/scalix/??/postgres/pgstartup.log and/or pgactivity.log

If postgres is working, and from the sound of it likely is, then I'd be checking disk IO next. The only other thing might be RAM, but usually that will just crash, not just be slow.

Also, do try the iostat thing, and see what kind of load is on the hard drives.

[brian.schnautz]

Valerion,
What do I do if there is already a user on my fresh install of centos 5.4 with the uid, gid of my existing scalix user?
old server: scalix:x:100:103:Scalix User:/var/opt/scalix:/bin/true
new server: avahi-autoipd:x:100:104:avahi-autoipd:/var/lib/avahi-autoipd:/sbin/nologin

Personally I would just migrate the mailstore over and let the installer handle the upgrade. Then you don't have to redo a lot of the configuration files in the system. I've had very little bad experiences, as long as you're careful.

1) Install the new system with the same hostname and IP, and create a Scalix user with the same UID and GID. Also create a Scalix group (check the live server first)
2) Copy /var/opt/scalix over, preserving permissions and ownership
3) Install Scalix and let it upgrade the mailstore
4) Test carefully

Also, be careful of VM's. For low-load systems it's not an issue, but disk IO is the first bottleneck you'll encouter. If you run Scalix in a VM over something like FC SCSI or iSCSI it'll mostly likely run very well. If you're running the mailstore inside the VM itself, you'll likely get issues with more than a handful of users, though.

[Valerion]

In that case you need to make sure you preserve the user and group names. Tar (normally) will do this correctly for you, and I have managed it with rsync as well, as long as it's not running chrooted (the default). Try to extract the archive and check the username and groupname.

If all else fails, you can later on correct the permissions with omcheck.