Warning - InRow RD: Excessive Compressor Cycling exists.

[Anonymous user]

HI Everyone,

We have a new Datacenter, which consists in part of 2 CRAC units set up in a group - HACS.  I believe that gives us about 16 tonnes of cooling, which is over-sized for the room we are in.  That however was understood when we went into the project.  We had the warning shown above, when the system first went into production - it was tuned, etc. and the alert went away.  We also added more heat load the room, which certainly helped.  It ran stable and quiet for about 5 months.

However all of a sudden we are getting this warning every day - multiple times a day.  Very strange as our heat load is higher than when the system went in.  The room is isolated from the building HVAC, so nothing has changed RE: climate in the room.

Any suggestions on what to look for.  When we called Schneider they advised that our load in the room just was not high enough - but that doesn't work for me, because the room ran fine for at least 5 months.

I am also wondering if our setup is correct - the group setup is HACS - but we have no ducting.  If I understand correctly the group should be set as In Row?  I can send pictures if it helps to see our room.  I am not really sure what happens if we make the change to the group, etc.

At any rate, if anyone has some suggestions we can look at, would be greatly appreciated.

[BillP]

Hello.

I am assuming that you have the ACRD5xx units in this configuration. The excessive compressor cycling alarm means that the unit is kicking on and off to often. This can shorten the life of the compressor. Like a car engine it is better to get the compressor up and running and let it run for a while instead of turning it on and off for short spans. The most common reason for short cycling is low load. The ACRD5xx needs about 10KW (about 3 tonnes) of cooling per unit to run at minimum load.

Since the units are in a HACS mode, do you have a hot isle containment system in place? If you do are there any gaps in your server racks? If there are gaps we are letting heat escape through them instead of going through the cooling units. This will cause the low load issues as there is not enough heat getting to the cooling units to cool. I would first try to seal up the gaps in the servers.

If there is no hot isle containment system in place or all the gaps are sealed, we can try a different mode. Put the unit into inrow but make sure the rack inlet sensors are placed in front of the racks of servers. The unit will use these sensors to cool, so if they are not properly placed there will be issues.

Try these steps first and if the units are still short cycling we can dial in the unit some more. Thank you

[Anonymous user]

Ok.  I know we have some blanking plates to install, so I will take a look at getting them in.

Looking at the units, I am seeing a Cold Air Demand of 0 KW and on the second unit 0 KW after a cooling cycle completes.  When cooling cycle kicks in, that demand is up to 30+ KW - after a few minutes it is down to 15 and/or 16  (I am going to see if I can time it tomorrow - how long it goes from the high number, to 0).  With that number, it appears we are meeting the bare minimum, even without some of the blanking plates.  That would make sense, as we were running without any alerts/warnings, up until 2 weeks ago.  At which point we started to receive the excessive cycling warning.

The other question, that has sort of bugged me after the system went in.  It is set to HACS - but we really don't have any hot air containment, other than the blanking plates in the racks.  We have nothing to the sides/top of the racks that seal in the hot air - there no true hot air containment in an aisle.

If I switch it to In Row, what change am I really making, and what should I watch for?  I am not an A/C guy, so I apologize for my ignorance on this matter.  (All I did know is I wanted the APC cooling system, as it made more sense to me in terms of cooling).

Our configuration is set like this - 2 Racks (minor data equipment - not a lot of heat) and then one CRAC.  Then 2 Data Racks (majority of the heat is generated here), and then the 2nd CRAC.  After that the other components of the infrastructure system (Power connections and Modular UPS racks).  When it first went in, we ran one CRAC only - but cooling did not get to the far data rack in sufficient quantity - so we switched to both running - which improved the room greatly.

[BillP]

Hello. If there is no hot isle containment system in place we do not want to use the HACS mode. We would want to either put the units to Inrow or RACS mode. In Inrow mode the unit will run off the rack inlet temp sensors. Make sure the sensor are properly placed in the front of the server racks and are not sitting on top of a server.

With the setup you described with your racks, make sure you have a sensor in front of each rack (there are 3 sensors per AC unit) and keep a close eye on the temperatures at the rack inlet and supply air for the first few days after changing the mode.

Thank you

[Anonymous user]

We have sensors installed on the midway point of each door in the racks.  Both CRAC units have 3 cables plugged into the input for Temperature sensors.  Are we losing anything by switching from HACS to In Row?  Our biggest issue is redundancy - we went through several A/C failures before we moved to the APC system (it was the impetus to get it installed).

I will try a switch today and monitor.  Interestingly the compressor cycling message stopped last night at about 2AM, and has not happened again.  This is what is confusing - why it all of sudden appeared.

[BillP · Accepted solution]

Recently you requested personal assistance from the Schneider Electric Customer Care Center through our on-line support center supporting APC products. Below is a summary of your request and our response.

We will assume your issue has been resolved if we do not hear from you within 21 days.

Thank you for allowing us to be of service to you.

[Anonymous user]

We are monitoring these devices via Solarwinds, which includes the Unit Cool Demand.  Decided to run some comparison numbers, as we could not explain why the alerts would all of sudden start up when server density had actually increased from initial setup.

I had a suspicion it was due to Server Load, and it appears I was correct.  During March Break, the Cool Demand was essentially cut in half, from normal operations.  We had a number of staff off on holidays - and our servers were probably running with much less demand - which in turn means less heat.  We have yet to have a compressor alert/warning since the break ended.

We have switched our units from HACS to Inrow, and will monitor for a week.  Since we have nothing to actually contain the heat, it makes sense that this configuration be used.  Are there things that I should watch for, other than higher inlet temperatures?

[BillP]

Recently you requested personal assistance from the Schneider Electric Customer Care Center through our on-line support center supporting APC products. Below is a summary of your request and our response.

We will assume your issue has been resolved if we do not hear from you within 21 days.

Thank you for allowing us to be of service to you.

[Anonymous user]

Well, we switched to In Row.  Did not notice anything during the day, but towards the evening started getting Compressor Cycle warnings.  Switched back to HACS - warnings went away.

[BillP]

Hello.

At this point I believe it would be best to gather the data and event logs out of the unit so we can better see what is happening when the unit is throwing this alarm. That way we have a few more tools to diagnose this. Here is what I would recommend:

1: In the web interface of the unit, go to the logs tab, then under data select the intervals option. Set this to one minute. That way we have a higher resolution of data.

2: Let the unit run a while until it hits a fit of compressor cycling alarms. If these issues do come in waves, let the event pass and then download the log files. If you are unsure of how to do this, please follow this link. If this does not work, please go to our knowledge base and search for answer ID FA156977.

3: Once you have the files, please either upload it in a way we can download the files (assuming you have hosting of some kind), or just let us know via this system or our email help services.

After that, we should be able to better diagnose what is happening during those times, and go from there.