eunsungc
commented
9 years ago Added host, prog, and volume fields. 1) host: get by running 'hostname --ip-address'; dest_ip var in GT6 is in fact the other end's ip address. client_ip looks like ip of a machine where globes-url-copy is run. 2) prog is set to 'globes-url-copy' 3) volume: set to mount point where the file is located.
On 9/2/15 4:51 PM, Raj Kettimuthu wrote: Regarding the clarifications: 1) Confirm
Fantastic.
2) I think you are processing the logs from the read-only servers
I figured as much and I'll build in similar logic to the netlogger logs on how we generate metadata depending on if it's a RETR or STOR transaction.
3) Failed transfers are not logged
Great. We were logging failure conditions from the netlogger logs but if this isn't in play with these ones, that's good to know.
Thanks so much for the feedback!
Cheers all,
Monte
Hey Brian,
Andy had our meeting and went through the new logs and mostly got things figured out. Hoped to avoid making any changes to the json logs, but we need at least one change. Also need some clarification on some points.
1) The main "show stopper" class addition we need to the json logs - the HOST value from the netlogger logs. The DEST is already in the json, but we absolutely need the HOST value.
1a) While they are adding that HOST value, if possible, it would be preferable if they make the value an IP address (v4 or v6) rather than the DNS version.
2) A couple of "nice to haves" while we're asking for things since we were capturing them from the netlogger logs:
Those are values that we are currently capturing, so it would be cool if they could add those in as well. But if not, it isn't a big show stopper or anything.
1) By our reading, the event_type "Transfer-End" corresponds to the 226 FTP code (closing connection) in the netlogger logs. We are only processing those entries. To that point, it looks like the event_type "Perf-Marker" entries are some sort of "in progress" logging.
So by our reading, it looks like the Transfer-End are the only entries we need to process. Confirm/Deny?
2) In the logs I grabbed, the only cmd_type (corresponds to TYPE in the netlogger logs) is RETR - there are no STOR commands. Is this merely a quirk of the logs I happened to grab, or are only those types being logged?
3) Are failures being logged to these logs? If so, is there an event_type that corresponds to a failure condition? Or is there some other way that we can detect a failure condition that we should look for so we can log the failures?
Cheers,
Monte