As if on cue after our report about its silence, the 3000-L mailing list and newsgroup sprang back to life over the weekend. One problem solved by the 415 readers was how to identify if a store to disk backup is a LZW compressed backup file. A Tracy Johnson report also confirmed that a LISTF,2 can report the time of each LISTF, by writing a specialized job.
Meanwhile, a 15-year-old HP 3000 with network connection troubles got advice from the newsgroup's readers. A Series 969 running MPE/iX 6.0 would not be the first thing you'd choose for interfacing to an internal website. But when a 3000 has data that a user needs over the Web, the server is the place to go.
Trouble started to surface when clients access a webpage which then opens a telnet session with the 3000, grabs the info, and then returns the data to the webpage.
We’ve been getting more and more errors over the last year, culminating in non-stop Could not initialize data in path with TCP, which then blocked anyone accessing us through our webserver. We’ve tried many changes but cannot seem to get past this.
When it locks up, the HP 3000 keeps running but won’t accept any new sessions. Which means our clients can’t run searches. Which is very bad for us. Sometimes we can stopnet and startnet and it will work for a while, but then the errors start again. Eventually, we have to coolstart to be able to have clients log in.
Mark Ranft suggested "If they are already running in the CS queue, here is the likely cause. Is there some new monitoring in place? If so, it may not be behaving well on the network.
What happens is someone uses a telnet or ftp connection to monitor whether the network is up on the 3000. They send the SYN, the hp3000 answers with the SYN ACK, and then the 3000 receives a reset before the connection is complete. This handshake sequence causes this exact error.
Also inetd and other HP3000 networking improved greatly in later versions of MPE/iX. If they must stay on 6.0, they should at least be sure to install the latest patches.
A third party support company served this troubled user until the support vendor folded, "and the only options we found weren't affordable." Getting the 3000 back up will trigger a revisit of that situation. Any server with critical customer data on it—and doesn't have a support vendor—relies on the largesse of the 3000 volunteers of 3000-L. That mailing list did go without a new message for more than a month, a troublesome response time for a critical server.