SMTP issue on a stretch cluster

[Kalin Tzenov]

Hello,

I have a 4 node remote cluster - 2 nodes on each site, with 2 system disks - one per site. Most of disk drives are shadowed accross the sites, however, we have few local shadow sets, shared between the local nodes on each site.

Here's my SMTP queues for the cluster. SCP001 and SCP002 are the nodes on one site, and SCP006 and SCP007 are the nodes on the other site.

$ sh que tcpip*/fu                                                         
Generic server queue TCPIP$SMTP                                            
  /GENERIC=(TCPIP$SMTP_SCP001_1,TCPIP$SMTP_SCP002_1,TCPIP$SMTP_SCP007_1,   
  TCPIP$SMTP_SCP006_1,TCPIP$SMTP_SCQML1_1,TCPIP$SMTP_SCQMF1_1)             
  /OWNER=[IST,SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S)                        
                                                                           
Server queue TCPIP$SMTP_SCP001_1, idle, on SCP001::, mounted form DEFAULT  
  /BASE_PRIORITY=4 /DEFAULT=(FEED,FORM=DEFAULT) /OWNER=[IST,SYSTEM]        
  /PROCESSOR=TCPIP$SMTP_SYMBIONT /PROTECTION=(S:M,O:D,G:R,W:S)             
                                                                           
Server queue TCPIP$SMTP_SCP002_1, idle, on SCP002::, mounted form DEFAULT  
  /BASE_PRIORITY=4 /DEFAULT=(FEED,FORM=DEFAULT) /OWNER=[IST,SYSTEM]        
  /PROCESSOR=TCPIP$SMTP_SYMBIONT /PROTECTION=(S:M,O:D,G:R,W:S)             
                                                                           
Server queue TCPIP$SMTP_SCP006_1, idle, on SCP006::, mounted form DEFAULT  
  /BASE_PRIORITY=4 /DEFAULT=(FEED,FORM=DEFAULT) /OWNER=[IST,SYSTEM]        
  /PROCESSOR=TCPIP$SMTP_SYMBIONT /PROTECTION=(S:M,O:D,G:R,W:S)             
                                                                           
Server queue TCPIP$SMTP_SCP007_1, idle, on SCP007::, mounted form DEFAULT  
  /BASE_PRIORITY=4 /DEFAULT=(FEED,FORM=DEFAULT) /OWNER=[IST,SYSTEM]        
  /PROCESSOR=TCPIP$SMTP_SYMBIONT /PROTECTION=(S:M,O:D,G:R,W:S)     

When sending a SMTP mail, the control file gets submitted to the first execution queue, in this case TCPIP$SMTP_SCP001_1. If I send the mail from SCP001 or SCP002 it works fine, however, if I send it from SCP006 or SCP007, the queue TCPIP$SMTP_SCP001_1 can't access the control file as it is on another disk which is not mounted on SCP001 (in mt case the system disk for the other site).

A workaround is to mount the remote system disk on the node where the first execution queue is, but I don't want to do this.

This issue wouldn't exist if I had a common system disk and all my drives were mounted on all nodes of the cluster, but unfortunately it's not an option.

Is there a way to force SMTP to submit the control file to the local node execution queue (in my case TCPIP$SMTP_SCP006_1)?

Thanks,

[Volker Halle · Accepted solution]

Kalin,

it might be best to try to configure 2 TCPIP$SMTP* generic queues, one on each site of this multi-site cluster. You seem to be running TCPIP V5.7, so have a look at SYS$SPECIFIC:[TCPIP$SMTP]TCPIP$SMTP.CONF file.

Volker.

[Kalin Tzenov]

Hi Volker,

Thank you for your advice. Worked like a charm :-)

Thanks again and have a good one!