Automatic graceful shutdown on multiple servers -- clarification needed

[Anonymous user]

I'm trying to get automatic graceful shutdown set up for the servers connected to my Smart-UPS using PowerChute. As far as I can tell, the procedure is:

# Install Agent on each server plugged into the UPS.
# Install Server on one server plugged into the UPS.
# Connect the Server server to the UPS via USB
# Add the network address of each Agent server to Server.
# When the UPS loses power, the Server server picks up on it, and sends a message to each Agent server, directing it to shut down.

This seems reasonable, but according to this support page, I need to buy additional hardware to get more than one server to shut down multiple servers when power is lost. Could someone please clarify what I need to do to get automatic graceful shutdown working? Thanks in advance.

[Anonymous user]

I'm trying to get automatic graceful shutdown set up for the servers connected to my Smart-UPS using PowerChute. As far as I can tell, the procedure is:

# Install Agent on each server plugged into the UPS.
# Install Server on one server plugged into the UPS.
# Connect the Server server to the UPS via USB
# Add the network address of each Agent server to Server.
# When the UPS loses power, the Server server picks up on it, and sends a message to each Agent server, directing it to shut down.

This seems reasonable, but according to this support page, I need to buy additional hardware to get more than one server to shut down multiple servers when power is lost. Could someone please clarify what I need to do to get automatic graceful shutdown working? Thanks in advance.

[TheNotoriousKMP_apc]

Joe,

Low battery state and low battery duration are the same thing. It wouldn't necessarily say Low Battery State, I was moreso talking about the state of the UPS being in Low Battery.

[BillP]

Greetings,

You're on the right path. You would need to buy an additional piece of hardware to allow for the Smart-UPS to communicate with more than one PC.

*1-3 Computers via Serial*: [AP9607CB Interface Expander Card|http://www.apc.com/resource/include/techspec_index.cfm?base_sku=AP9607CB]. This card would install into the bay on the back of your Smart-UPS. The card provides two simple signalling ports (cables included). This solution would allow up to 3 total PCs to be gracefully shut down (1 Smart signalling via built-in serial or USB, and 2 Simple Signalling). You would continue to use PowerChute Business Edition Basic and install the Agent component on each PC connected via serial/USB to the Smart-UPS. Any PC connected via the simple signaling ports on the expander card would need to be set up for "simple signaling" when installing the Agent component.

*1-8 Computers via Serial*: [AP9207 Share-UPS Interface Expander|http://www.apc.com/resource/include/techspec_index.cfm?base_sku=AP9207]. This device plugs into your Smart-UPS' built-in serial port and provides connectivity for up to 8 computers to be shut down via PowerChute Business Edition agents. Cables are not included. You would use this solution if you need to shutdown more than 3 computers. In my opinion, if you need to shut down more than 3 computers, you would want to go with this next option...

*1+ Computers via Network*: [Network Management Card|http://www.apc.com/resource/include/techspec_index.cfm?base_sku=AP9617] (now standard equipment in many Smart-UPS). By installing a network management card in your Smart-UPS accessory bay, you give the UPS a presence on your network. It would be reachable by several methods such as HTTP and TELNET, or their secure counterparts HTTPS and SSH. With this card installed, you would no longer use any USB or serial cables for UPS to PC communication. Instead, you would simply use your existing ethernet network and [PowerChute Network Shutdown|http://www.apc.com/resource/include/techspec_index.cfm?base_sku=SFPCNS223&tab=Software]. Unlike with the AP9607CB Expander Card, the AP9617 Network Card does not limit you to 3 computers. You can shut down as many computers as your UPS can safely support.

PCNS is similar to the PowerChute Business Edition Agent component. The PCNS service runs in the background and maintains communication with the UPS's network card. By default, when the UPS reaches the low-battery state, the PCs running PCNS will begin to gracefully shutdown. There are many more customization options for shutting down computers in the PCNS web interface.

Do either of these options seem to be good match for your configuration? Please let us know if we can assist you further.

[Anonymous user]

Thanks, everyone!

I'll have eight devices plugged into each UPS, so making it a network device sounds like the best option. If I do this, do I still have to install PowerChute Server on a system, or are the Agents smart enough to monitor the UPS and shut it down when necessary? And can I set different behavior in the Agent for loss of network connectivity to the UPS, as opposed to a report of power loss?

If I install the network card in the UPS, what are the SNMP monitoring options? Can I monitor the UPS directly, or does the server with PowerChute Server installed act as a SNMP front-end for the UPS?

[BillP]

Good choice with the network card option.

You will now be using PowerChute Network Shutdown exclusively. Uninstall any PowerChute Business Edition components anywhere on the computers that will be part of this solution. There is only one component in PCNS: Agent. Simply install the PCNS agent onto any of the computers that need the graceful shutdown. During the install, you will pair up the comptuer with the network card in the UPS.

Yes, you can configure shutdowns and other activities (ie. command files, notification) based on many UPS events, such as loss of comm. This is all done by logging into the PCNS web interface located at http://:3052.

With the network management card, you have full SNMP access directly from the UPS. Feel free to use whatever SNMP compatible tool you like for traps, sets, and gets. Again, there is no "server" component in PCNS -- just the Network Card in the UPS and the PCNS clients on your network. Management is via IP to either the UPS' network card or the servers running PCNS.

[Anonymous user]

i have a brand spanking new smart-ups rt 6000, and a brand spanking new ap9619 network card. i'm trying to justify buying this new equipment by instituting graceful shutdowns.
i canna find anything in more than 6 hours of searching the apc database about actually testing graceful shutdowns. the installation was fairly easy, and i could as well install this onto more than 7 different flavors of OS, but as far as actually testing the equipment and seeing it work, alls i can find is something like "when an event occurs the server receives a notification and triggers the graceful shutdown".
to be fair, the configuration page is chock full of configuration stuff, but actually testing the system (and thereby justifying its existence) is gosh darn hard to find. can you point me to a pdf that talks about using the product, and in particular about graceful shutdown tests? plug and pray just won't work for me in this case.

[TheNotoriousKMP_apc]

Joe,

Unfortunately, there's no documentation outlining testing conditions for graceful shutdown between PCNS and an NMC.

The easiest thing to do would be to configure the PCNS client to do a shutdown for UPS: Communication Lost, and just pull the plug on the network jack on either the NMC or the PCNS client side.

Then you should be able to watch the graceful shutdown process.

[Anonymous user]

Excellent suggestion; thank you very much. i'll try it tomorrow morning before the kids come into work. - i'm also keen to try the low battery event, but the description of what a low battery actually means is a bit mystifying. is there a "layman's terms" description of a low battery? i mean, i can set it for 20 minutes, or 5 minutes, or even 2 minutes. i find it easier to understand the technical definition once i understand the concept. - what secretly worries me is that everything i read says that a "low battery" condition forces the servers into a graceful shutdown. this is good, but i'd like to test it, and i think i could test it if i could figure out what the time settings mean for a low battery.

the unit itself looks like it is doing just what we needed, so we are pretty happy. testing is my last step. thanks again for the handy suggestion on "communication lost". pretty easy test, i think.

[TheNotoriousKMP_apc]

Low Battery duration is basically the value at which you set the UPS to initiate a graceful shutdown for any devices being run by PCNS that are still online. The reason for different values is that some servers take longer than others to shutdown. Why it's set to 2 minutes by default, I do not know. I have yet to encounter a server that can shutdown in 2 minutes or less.

[Anonymous user]

Thank you again. The test worked. Because I was ascared, I set these three:
Communication: Lost while on Battery
Communication: Management Card cannot communicate with the UPS
Communication: PowerChute cannot communicate with the Management Card
I didn't think our scheme of pulling the network cable would kick off "Lost while on Battery" because we'd be on building power at that time.
I didn't understand how the Management Card cannot communicate with the UPS unless the MC is physically (or logically) broken.
So I think what you and I were aiming at was "PowerChute cannot communicate with the Management Card." Right?

I'm still not quite clear on "Low Battery". I think I'd like to set it way way way high, so that I could test it. (Ours is 6KVA, so it's hard to get it to run down.)

And, I'm having trouble with IE7 and certificates for 127.0.0.1, although I can still talk to the agent. IE6 seems to be OK. I'm uncomfortable with that, but it doesn't surprise me that IE7 would give me yet another problem.

IS there a chat area for this level of UPS product? All of this would be easier in a chat session.
Thanks very much again.
Joe

[TheNotoriousKMP_apc]

Joe,

Here's what I'll do for you. There is a chat support for this level, however, you may want to talk to one of us (moderators) instead. That being said.....

If you still have any outstanding issues or questions and don't want to wait for the turnaround on a forum post, log into the Chat agent tomorrow between 8 and 430, thursday between 9 and 530, or friday between 12 and 8pm, all EST. And request to speak with me. They'll transfer you over to me. I don't necessarily do our livechat on our page, I do escalation chat (internal) for APC Reps. However, I use the same chat program. Just make sure you give them your name and have them ping me directly.

Tell them to ask for Kevin P.

[Anonymous user]

thank you. i am definitely an American living in SuZhou, and we are 12 hours different between ET. so that is part of my condition.
the only thing Chinese about this project is the business location and the people; my servers are US english, my UPS is US english (mostly), and my software is all all all US english. my current situation is remarkably better than the other 6KVA UPS that was here for two weeks and the other supplier golden service team could never get it out of "internal test" mode. gosh.
we are making headway. most all of my issues could have been solved with a chat session, but it is gosh darn difficult to make that happen. so we are using the poor man's chat, which is back-and-forth posting. gosh.
the UPS is working well. i just need to understand the standard configuration, and be able to test it. to wit, i need to see what it does when building power goes out, which i tested yesterday. i need to see what it does when the battery loses connectivity, or when it enters "low battery" state, because that might be a better solution that "on battery". but i haven't figured out the meaning of those settings, and then testing "low battery" might be hard for me to do.
and of course the certificate issue with IE7, but that seems to be only a detail.

so thank you for your patience, and support. you can sense that a chat session is what i really needed, but we are nearly worlds apart. so maybe your schedule will fit with mine, and i'll be ok. meanwhile, i'll continue to do my homework on "low battery", and chase IE7 in my ample spare time.
thanks, Kevin!

[TheNotoriousKMP_apc]

Low battery state is just "when the UPS has X runtime remaining, send a broadcast for everything with PCNS to shutdown immediately" . If you're going to test it, be ready for the incessant beeping when it reaches a low battery state.

I would recommend sticking with the After X amount of time On Battery, mainly because you can prevent a hard shutdown from consecutive power outages.

[Anonymous user]

thanks again. now i've tried to find "low battery state" and can not find that. i expect that it is on the UPS (and not on the servers), but I can't find it. i can find "Low-Battery Duration", but not low battery state.
What I want, of course, is to find the setting that says: "this is where you are when a low battery state or condition occurs". Then I could spin that setting up as high as i can, run the battery down to that point, and watch my servers shut down gracefully. Otherwise, i'm taking it on blind faith that a broadcast is sent for everything with a PCNS. Gosh.
So each day i get closer. it is a little wretched to have to remember where i was yesterday when i start studying again today, but at least the servers are on the UPS and they are ready for "UPS: On battery" which was our chief worry when we started looking for a solution.
if you can point me to a PDF that talks about "low-battery" i will study it and implement it.
thanks again. . .

[Anonymous user]

OK. Thank you. I am inferring "Communication: PowerChute CCWTMC" and "UPS: On Battery" are server-side settings. I'm also inferring that "Low Battery" is a UPS-side setting.
So one can configure one's servers for the Communication and On-Battery pieces, and the "Low Battery" config is on the UPS.
AND, importantly, the "Low Battery" config trumps server-side settings. Once the "Low Battery" event occurs, the UPS will send a broadcast for everything with PCNS to shutdown immediately.
And of course I would want to test that. AND understand the ramifcations of knowing what constitutes a "Low Battery" event; for example, it is probably different if you have 1 server than if you have 20 servers sitting on the UPS.

See what you think about the above, and I'll try to find a way to test "Low Battery", maybe on the weekend. Again, if I'm missing some basic documentation that walks me through all of this, let me know and I will read it.
Thanks again, Kevin.
Joe

[TheNotoriousKMP_apc]

Here's what you have to take into consideration...

A server load is considered a Fluctuating Load, therefore, the runtime and load % you see when you log into the NMC may differ when the UPS starts operating on battery - making it extremely tough to determine what the value for "After X minutes on battery" is with multiple servers. Most times, if you're using multiple servers, you would use a Low Battery Duration equal to or above the longest server shutdown time.

Therefore, with one server, I'd say, if you're going to test it, do the following....

1. Set the low battery signal time for the highest possible value.
2. Start running a scan or something on the server to increase the load and inversely decrease the runtime faster.

Then watch what happens when it hits Low Battery state.

[Anonymous user · Accepted solution]

Thanks, NK. I will try to devise some sort of test for that.
Here's my gut-wrench: the settings that I have built (communication lost, for example), are easy to test (thanks to your suggestion). I wait until 6 pm, pull the network cable, and three minutes later the servers are shutting down.
But we bought the UPS graceful shutdown expressly for the times when a carbon unit is NOT at the office, like 3 am. Take away my building power at 3 pm, and I can gracefully shut down the servers. But take it away at 3 am, and I am hosed without a graceful shutdown. (After battery time, of course).
So your test is a good idea. I'll work on it. I just only wish there were some easier 5,000 ft documentation on automatic shutdowns. I surely didn't buy the APC unit because there are lots of English speaking tech-sales here in SuZhou. There are none. So I’m on my own.
You can close this thread, or whatever you call it when a user is happy. I think we are set now, and I’ll find the way to test low-battery; your explanations have all been very helpful. (And patient!) Thanks!