Invalid FTP Handle

[Paul DiLorenzo]

I'm running HP-UX 11.0 on an L-Class server. This strange problem began happening last week and has been consistent every night. Between midnight & 1:20 AM, FTP jobs fail with the message "Invalid File Handle" (this message is seen in the SAP application). No other o/s level messages are found. We can run this same FTP any other time of the day and it works fine. I see nothing in syslog, dmesg, or netstat listings to indicate anything. Where else can I look for diagnostic infornation? This is the only box this happens on.

Thanks,

Paul DiLorenzo

[Pete Randall]

Paul,

I didn't have time to review this but a search turned up these results:

http://us-support.external.hp.com/emse/bin/doc.pl/sid=15938ceb0e8f5e5ae0?todo=search&searchtext=invalid+ftp+handle&x=27&y=9&searchcriteria=exactphrase&searchtype=SEARCH_TECH_DOCS&searchtype=SEARCH_PATCH&searchcategory=HPUX_1100&searchtype=SEARCH_MANUAL&searchtype=SEARCH_FORUMS&searchcategory=ALL&rn=25&presort=rank


Pete

[Massimo Bianchi]

Hi,
are you using sapftp ?
There are a lot of notes in sap regarding similar issues.

Can you enable the cpic trace for the night, to have more information ?

Is it always the same job that fail? It could be a damaged/corrupted file. Have you tried looking at that file ?

Massimo

[Krishna Prasad]

I remember seeing this error when our Network Firewall rules changed.

It started working again when we used FTP in passive mode.

What is strange in your case is that it depends on the time of day? I would think the rule would be static? Also, I don't know if you are going through a firewall?


open someplace.com
passive

[Ron Kinner]

My bet would be someone is backing up the target machine (or running some other kind of CRON job) starting at midnight and the process is causing the problem.

Ron

[Paul DiLorenzo]

Thanks for the replies! We are running sapftp and I did just start a trace through SAP. However, I just found out that it's not only the ftp jobs that are affected in this timeframe. We also run MQ-series tranfers which fail with "Unable to connect to target host" message. This is also occuring between the same midnight-1:20AM time. Again, running these same jobs at any other time is successful. This is the only server experiencing this.

Thanks,

Paul DiLorenzo

[Ron Kinner]

OK It's not the target. It's your machine that is doing the backup at midnight. You are using the same network card for the backup as for your other connectivity and the backup is hogging the LAN so your other stuff times out.

Or there is another process running on a machine on the LAN which is hogging the 10 half entrance to the router. I had this when we first tried to go to a new switch. Most of the other machines were 10/half but one big talker was 100 full and every time he did his once an hour thing he completely swamped the input to the router so that nobody else could get through and everyone else thought the network was down.

Ron

[Gerardo J Oliva Galvan]

We have a similar case. That's what we found.
Check the bandwith of the net (10, 100, 1000), full duplex, half duplex, and validate that the speed configured on the network cards matches the speed given to you by networking-dept. It it does not match can cause you trouble.

If someone is running a backup (omniback, legato, oracle, etc) on the target machine it can take the whole bandwith, so watch the schedules.

In my case we found that not only the port for the server has to be validated, but also the ports next to it (some networking devices, may be impacted by the network usage on the ports adjacent to your port), this was tha case we found with cisco 100-halfduplex, we have to change the 100-halpdulplex cards to 10/100 fullduplex.

[Massimo Bianchi]

Hi,
in addition to other hint, if it were possible i would shutdown MQseries during the backup.

We had a cluster with autosys, mqseries, connectdirect, connecenterprise, oracle, sap mcgs and if we do not stop all of the the offline seldom works.

Mq series is very demanding on semaphores and shared memory and tcp/ip connections.

Massimo

[Paul DiLorenzo]

Thank you all for your suggestions. As it turned out, I found a message buried in an SAP trace that mentioned "insufficient space for fork". I increased swap space, and both the FTP and MQ jobs began running successfuly again. I had assumed it was a network issue because of the error messages "Invalid FTP handle" and "unable to reach target host". The jobs were failing before they made it to the netwok.

Thanks again