IBM Integration Community Come for answers. Stay for best practices. All we’re missing is you. Join / Log in Ask a question
Our organization has a cluster of 4 DataPowers that connect to a FTP server and poll different folders for file transfer
1. 2 DataPowers in data center MDC
2. 2 DataPowers in data center PDC
We have about 40 MPGW services with pollers that use FTPS protocol to connect to a FTP server to poll different ftp server folder for file and transfer the file to a agency. As we want high availability, we deploy each service is deployed to 4 DataPwoers (on 2 data centers). The polling interval is 1 minutes. So there will be about 40x4=160 poling to the FTP server in about 1 minute.
We see a lot of read errors (could not allocate ssh connection, failed to open file, and etc.) each day maybe because too many services (160) to try to connect to the FTP server in one minute. . I would like to ask DataPower forrum a question, is there a way that if one DataPower (out of 4 DataPowers) polls and detects a file, other 3 DataPowers will know it and not try to connect to FTP server to poll the FTP server folder. As we want to keep high availability, we don't want to deploy a service just on 1 DataPower.
I'm afraid there isn't a way to make "cluster" members aware of each other hen using pollers. We had sort of a same problem having a lot of error messages from pollers saturating the logs and one idea we had was to selectively increase the polling interval of "cluster members". So basically we would have had one appliance that polls more frequently (eg. 60 secs) and other members then have longer periods in which they poll (eg. 120 secs, 180 secs....) The interval is controlled by deployment policies during deployment time.
Don't know if this idea is feasible for you, but just throwing out some ideas...
I like your idea. I would like to discuss with my manager and would like to try it in our change request windows. I shall update here after I try this idea.
Thanks a lot for your help.