Plant Connectivity: IsConnected returned FALSE - warning or error?

[janbuchmann]

Hi there

We have a number of socket agents in PCo for TCP/IP source systems (e.g. weighing scales). These connections can drop for various reasons, which lead to "IsConnected returned FALSE" events. What puzzles me is that these are only logged as warnings, not errors. Once the maximum number of retries fails, the agent switches to a faulty state. But shouldn't this (at least) lead to an error?

Some additional background: we do not have MII in our landscape, thus I am sort of limited in terms of active monitoring and was hoping to use SCOM. But: for this I would need event log entries with errors, not warnings. I initially set "Maximum Number of Retries" to something like 999, hoping that PCo will re-establish the TCP/IP socket as soon as the reconnect is successful. This does not seem to be the case. I therefore set the "Maximum Number of Retries" to 0, hoping to get errors which can trigger SCOM to send an alert and/or automatically restart the applicable windows service.

Any thoughts/suggestions?

Thanks & regards
Jan

[francois_bouchard]

Hi Steeve, do you think we must have a specific setup on the vlinx side ? We are at version 15.4.3.13, do you think an upgrade would be require since it's not working.

Regards,

Francois

[steve_stubbs]

Hi Francois,

We have seen issues with the VLinx device losing its network connection and having to be power-cycled to reset. This would also cause the Connnect=false issue in PCo. Our VLinx are set up for DHCP; DHCP Server reserves a fixed IP address by the VLinx MAC address. We have not been able to determine what causes the loss of network connection but it appears to be somehwere between VLinx and PCo.

I will check the VLInx software version on our unit next week and update this comment with any additional details.

Regards, Steve

[francois_bouchard]

Hello Jan, I don't know if you fixed your problem and I would be curious to have your resolution. We have a similar situation here where we loose the connection for a moment with an IP addresss (Vlinx connected to a scale), but seem to not be able to reconnect properly all the time leading to the notification being in error.

Tks

Francois

[steve_stubbs]

Hi Francois,

I have PCo 15.5.4.4 and a VLinx Serial - Ethernet server. If I cycle power to the device, or unplug and re-connect the ethernet cable, PCo Socket Agent recovers without error.

When you set the Source Source to Check Connection for a Data Port, The Agent Instance writes a NULL (0x00) byte to the Socket it is connected to. If this fails, then the log event "IsConnected returned FALSE" is logged for the Agent Instance. If the Vlinx is power cycled or the network cable is unplugged the Agent Instance will log some connection errors and begin the re-connect sequence for the number of retries you have configured for the Source Agent.

Regards, Steve

[janbuchmann]

Hi steve.stubbs

Are there perhaps any recommendations/best practices with regard to 'Retry interval' and 'Number of retries'? I'd imagine retry interval > power cycle time, right?

Thanks

[steve_stubbs]

Hi Jan,

In a production environment I would set Max. Number of Retries to a large number; i.e., 9999. If the Max. Number of Retries is exceeded without a successful reconnect, the Socket Agent Instance mode will Fault and will need to be re-started.

I have always left the Retry Interval at 30 s. With a large number of Retries, this should not have a significant impact.

If you have a way in your environment to execute Web Service calls to the PCo ManagementHost Service (renamed Main Service in PCo 15.5), you can get the state of an Agent Instance, retrieve an Agent Instance Log, and restart the Agent Instance. It is also possible to do this completely within PCo but takes a bit of configuration to implement.

Regards, Steve

[janbuchmann]

Hi Steve

Thank you so much for your reply and your valuable input. This sounds very interesting: It is also possible to do this completely within PCo but takes a bit of configuration to implement. Do you perhaps have more information/guides for it?

Regards
Jan

[Kevin_Hunter]

Hi Francois,

I wrote a white paper a few years ago that show how to implement a "self healing agent" completely in PCo and this is what Steve is talking about. It will need modification to match your scenario but it will show you the basics of how you can achieve it.

If you send me an email (kevin.hunter@sap.com) i will send you a copy as its not available on any download sites.

Thanks

Kevin