New cloverleaf is sending TCP PSH flags, 2 receiveing system have problems

[J.S.]

Hi

Last month we upgraded to 2022.09 rev2 on a RH linux system 9.3. (from a 6.2 on RH 7)

When during the day, when there is more traffic, wireshark/tcpdump show that PSH flags are beeing sent, telling the receiver to push it right away to the application layer. There the HL7 message is nog yet complete so it is erroring because of not correct start/end characters of the message.

Anyone also encoutered this behaviour? Any settings known we can adjust?

Thanks..

Jessica

Regards
J.S.

Topic

[Charlie Bursell]

Must be large messages where multiple packets are received.  Should it be inherent on the receiving application to buffer those packets until a complete message is received? For example, assuming MLP HL7 data, Cloverleaf noting a packet with a start character of x0b (vertical Tab),  would buffer the message until 1c0d (File separator/Carriage Return)  is received

Am I missing something?

[J.S.]

Yes we use MLP, and yes, with MTU of 1500 also this is sometimes in multiple packets. But this is not different from the old system, sending the PSH is.

PSH means transmit to the application immediate, and that seems to be in case of our PACS system not a complete segment.

Maybe they could improve something, but the change on cloverleaf side triggers this behaviour, hence the question if we could change settings on that side making sure PSH is not sent. There are quite some tcp settings involved.

I’m not completely sure but I think MLP is already a layer up on tcp.

 

Regards
J.S.

[Robert Kersemakers]

Hi Jessica,

We just installed CL 2022.09.03 on a new RHEL9.3 cluster in Azure.
Our Dutch reseller (you know them/him…) told us that there was a problem with 2022.09 on RHEL9 regarding this PSH flag that is suddenly being send.

Problem occurs only with 2022.09 on RHEL9; on RHEL7 it worked without a problem. And only certain systems are affected; probably those that actually adhere to the PSH flag and try to process the data when the PSH flag is sent.

Call was made for Infor to check this out. Not a nice thing to know as we just installed on RHEL9. And we also have Sectra as PACS, which is one of the offending systems.
I can’t find anything quickly about this PSH flag and how to disable it. Will research some more.

Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands

[Sandy McQuay]

This is great information.  We’re just starting to plan our upgrade from 6.2.2 on RHEL 7.9 to 2022.09.03 on RHEL x.x (not sure yet, whatever is certified).  We also have Sectra PACS.

We’ll keep this in mind as we move forward.

[Marc Pleijers]

Hi all,

Infor has done research on this issue together with Redhat.

In RHEL 7 and 8 there was a bug in the TCPIP stack that was fixed in RHEL9 and now the TCIP stack is completely according to the specifications.

This fix adds the PSH flags in HL7 messages of a certain minimum size, when Cloverleaf sends them to Sectra and other systems. This is exactly according to the TCPIP specifications.

For HL7 messages the MLP protocol is used on top of the TCPIP protocol. Receiving system Sectra should  buffer the message until  according to MLP protocol 1c0d (File separator/Carriage Return)  is received. They should not start processing once they get a PSH flag inside the message body.

Conclusion is that this is not a Redhat 9 /Cloverleaf 2022.09 issue but is an MLP implementation problem at the receiving Sectra System. Sectra has been informed about this.

Kind regards,

Marc Pleijers

 

• This reply was modified 2 weeks, 4 days ago by Marc Pleijers.

Marc Pleijers
Senior Integration Consultant
Enovation BV
The Netherlands

[Author]

Replies