Unstable communicatiom with pCOWeb

[DCIM_Support]

Hi guys!

I have two Leonardo downflow CRAC units with pCOWeb installed.

Both demonstrate very unstable communication via SNMP, Modbus TCP and Web. It allow to log in to the web interface but when I click some tab, the web card have no response and browser show timeout error. And after some time (2-5 min) it back to live again. Sometimes I even can not log in, but sometimes it works fine for a short time and I can log in and switch between tabs, but in 5 - 7 minutes I'll lost communication again.

The issue is similar on both units.

Can anyone advice how to troubleshoot it?

Here is communication status graph from DCE:

(CID:123343223)

[DCIM_Support]

Hi Valentin,

If this is only happening to these 2 devices, it does sound like the devices themselves. Add to this the fact that you mention that " the web card have no response and browser show timeout error." it sounds like the card is either busy or the network is having issues. Looking at the communications graph is quite interesting too. The graph looks perfect...like the lost and regained comms happen at precisely regular intervals (likely due to polling frequency). 

Do you know if there's something else polling the devices? Have you looked at the logs for the devices? Are there other devices on the same subnet that works? If the devices is in a distant subnet, have you tried installing a VM and removing the device from this DCE and adding it to a DCE VM on that same subnet to rule out networking issues?

Steve

(CID:123343236)

[DCIM_Support]

Hi Steve!

Devices directly connected to our APC switch and polled by DCE. Other devices on this subnet (UPS and PDUs) works fine, so probably it's a not network related issue.

I'll try to connect unit to VM directly, of course.

I have no wide experience with pCOWeb. Usually we are use Modbus RTU + gateway option and all works fine. This time customer order web cards, but I can't to add it to DCE by SNMP (DCE recognize it as Linux workstation). Thats why I need to use Modbus TCP.

(CID:123343358)

[DCIM_Support]

pCOWEB can connect by SNMP and Modbus IP if you wish

(CID:123343600)

[DCIM_Support]

Hi Valentin,

pCOWeb cards are weird when it comes to SNMP. They re-use OIDs. The same OID sued for your cooling unit for one sensor may be a totally different sensor on another system. Additionally, they often do not define themselves which is necessary when discovering via SNMP. If you have multiple pCOWeb devices, they will often just use the first pCOWeb based DDF  DCE finds for them. Modbus it often the best option.

Are there multiple slave addresses connected? If you have multiple slave addresses there could be an issue where there's too much issue on the bus. Similarly, if other systems are monitoring the device, that may interfere with DCE polling the device.

I'm also wondering if the way DCE is polling the device may have an effect on it. I have seen some instances (not pCOWeb specifically though) where we may poll individual registers and a device will work fine but if we poll a block of registers, the device doesn't respond well. 

Steve

(CID:123343606)

[DCIM_Support]

As another suggestion, it may be an idea to validate the network port that the card it plugged into on the switch.  Ask the network team to have a check on the duplex option and port speed.

(CID:123343260)

[DCIM_Support]

Thanks Warren! I'll try to check this out.

(CID:123343359)

[DCIM_Support · Accepted solution]

Hi All!

I fix my isue. The reason was in incorrect network settings on pCOWeb card.

It have 3 Eth subintarfaces and Eth0 was configured for DHCP and Eth1 was configured for static address.

Removing DHCP and setting static address for Eth0 resolve the issue.

(CID:123344308)

[DCIM_Support]

This question is closed for comments. You're welcome to start a new topic if you have further comments on this issue.