nsq_to_file: malformed gzip data

[mreiferson]

nsq_to_file can corrupt output files when gzip is enabled.

See #677

[arduanov]

I'm affected this bug, about once a day nsq_to_file produces malformed gzip file.
Start parameters:

nsq_to_file --filename-format="<TOPIC>/<DATETIME>/data<REV>.json.gz" -datetime-format "ds=%d-%H-%M"  --skip-empty-files=true -gzip

[mreiferson]

@arduanov can you provide any additional information beyond what's already been discussed in #677?

[arduanov]

I found the cause of the bug. Service falling due to an error and the gzip files becomes unreadable

ERR    7 [***/nsq_to_file] error querying nsqlookupd (http://127.0.0.1:4161/lookup?topic=***) - Get http://127.0.0.1:4161/lookup?topic=***: dial tcp 127.0.0.1:4161: socket: too many open files

[ploxiln]

That could be fixed in #935

[mreiferson]

@arduanov thanks, can you share more of the logs?

[arduanov]

Here

Sep 03 12:39:38 srv nsq_to_file[64971]: 2017/09/03 12:39:38 INF   11 [***/nsq_to_file] querying nsqlookupd http://127.0.0.1:4161/lookup?topic=***
Sep 03 12:39:37 srv nsq_to_file[64971]: 2017/09/03 12:39:37 ERR    7 [***/nsq_to_file] error querying nsqlookupd (http://127.0.0.1:4161/lookup?topic=***) - Get http://127.0.0.1:4161/lookup?topic=***: dial tcp 127.0.0.1:4161: socket: too many open files
Sep 03 12:39:37 srv nsq_to_file[64971]: 2017/09/03 12:39:37 INF    7 [***/nsq_to_file] querying nsqlookupd http://127.0.0.1:4161/lookup?topic=***
Sep 03 12:39:37 srv nsq_to_file[64971]: 2017/09/03 12:39:37 ERROR: could not retrieve topic list: Failed to query any nsqlookupd: Get http://127.0.0.1:4161/topics: dial tcp 127.0.0.1:4161: socket: too many open files
Sep 03 12:39:37 srv nsq_to_file[64971]: 2017/09/03 12:39:37 syncing 65 records to disk

[mreiferson]

Now that #935 and #1110 have landed, and the fact that we f.Sync() before responding FIN to messages, I'm going to consider this (and the original issue) resolved.