Why then?

The latest email from you guys:

In order to upgrade the RAM to 4GB in the CP server and upgrade the RAM to 3 GB in the mail server, both servers (control panel and mail) will go off-line between 1pm and 3pm ET October 1, 2002, for approximately 30 minutes each. This should also make the control about a lot more stable.

We will also be bringing 4 new Windows servers on-line, along with a new mail server.

This may be a dumb question, but why purposely bring down the servers in the middle of the afternoon on the east coast of the US? I imagine most of your customers (and mine) are in the US... why not do this during "off" hours so as not to piss off any more customers who are already screaming because of all the downtime lately?

I do fully agree with this. I understand that you have customers from all over the world, but more of them are from the U.S. than any other country aren't they? Scheduled downtime always seems to happen in the afternoon in the U.S. and I've also often wondered why. =)

Downtime sometime during 1 - 3pm is going to cause complaints from my customers for sure, and after yesterday I don't know how understanding they will be.

[mresell]

I would be interested to know the answer to this as well. This is what I was talking about in another post. It is got to be incredibly difficult to host globally out of 1 central location. I hate to sound like a US snob, but hey I have to be concerned w/ this. Maybe most of their customers are not in US.....

[vonbrocklin]

I thought the EXACT the same thing. My first thought was "Cool! Matrix is letting us know about downtime!" - until I read the message, and discovered the downtime was smack in the middle of the day for most U.S. clients. What gives?

1pm is quickly approaching. Could we get some input on this admin? I'd hate to have to explain to my customers why their mail is down again in the middle of the business day for the second day in a row. Thanks for your help.

[Garreg]

At last something to my advantage .... er I'm GMT

Seriously - I guess its hard pleasing all of us most of the time.... seems lately that no matter what they do 'we' don't like it.
Upgrading the ram is surely only a once in a blue moon thing... and given the recent problems with the CP and mail - whats the better of the two evils - 30 mins downtime in the day or continued on line / offline ... all of the time until it's resolved?

[BCS]

If 30 minutes during the middle of the day will resolve some of the problems we've been having for the last month ... well worth it.

Last night the CP area was taking up to 2 minutes to display sections/pages. This morning I had customer complaints of not being able to access the CP at all (suspect timeouts) along with blank pages showing up on thier actual site. They couldn't use the TT feature at all.

Answered them with how this planned upgrade will help resolve this ... hope I didn't mislead them but fingers, and any other appendage I have, are crossed.

One customer ended the email with ....
"This is starting to be a problem."

What an understatement.

[Scroller]

Originally posted by Garreg
At last something to my advantage .... er I'm GMT

Seriously - I guess its hard pleasing all of us most of the time.... seems lately that no matter what they do 'we' don't like it.
Upgrading the ram is surely only a once in a blue moon thing... and given the recent problems with the CP and mail - whats the better of the two evils - 30 mins downtime in the day or continued on line / offline ... all of the time until it's resolved?

Same here. I've been holding back, trying not to defend the choice of timing since I actually benefit from it. I'm 12hrs ahead of ET. But seriously, 30 mins is a small sacrifice if at least some of the many problems now can be resolved.

My personal experience ... sometimes it's better to do server upgrades in the day instead of the middle of the night. At least, if anything goes wrong, you can get help/parts immediately. Just me 2 cents.

[hly]

...well, the best practice is to have some form of redundancy so there would not be any downtime at all --- no matter what time of day you do your maintenance. If one server goes down, another takes over, and vice versa.

[BCS]

Originally posted by hly ...well, the best practice is to have some form of redundancy so there would not be any downtime at all --- no matter what time of day you do your maintenance. If one server goes down, another takes over, and vice versa.

... and I wish I'd stop finding an odd pair of socks in the dryer

Redundancy usually works great with hardware problems/upgrades ... most times you swap the drives for maintenance but still end up with some downtime while the network internally propagates the system. 30 minutes is reasonable even if this is the case.

Downtimes will magnify on a quick server swap on a fully redundant system if the backbone route resolves are cached by the ISPs/networks accessing the servers. If they happened to get a "server not found" error it may take up to 12 hours for them to allow the cache to clear.

If there is any type of software upgrade/fix involved, redundancy can pretty much become a moot point.

... *%$^&$!, where's that other sock

[Garreg]

BCS:

and I wish I'd stop finding an odd pair of socks in the dryer

Same here ! But it never worries me cause I've got another pair just like them

(sorry a bit of track- just couldn't help myself).

[admin]

Hey guys,

I know the timing kind of sucks but i wanted this done ASAP to get the CP back to stable..

Just so everyone knows tuesday is our day to do most maintenance but we have never had to touch a server as much as this H-Sphere cluster so its never been a real problem till now..

I hope this fixes some of the problems. So far it seems to be stable.