marija icon indicating copy to clipboard operation
marija copied to clipboard

Unable to use marija with elasticsearch 5.0.2 over ssh tunnel

Open ckuethe opened this issue 9 years ago • 11 comments

I'm connecting to a remote Elasticsearch 5.0.2 instance over an SSH tunnel; the tunnel works because:

  • connecting to http://localhost:9200/ in a browser returns the proper version information
  • marija detects my indexes

Attempting to "use the refresh icon to refresh the list of available fields" or "add the fields [...] to use as nodes" gives an error

Error executing query: The connection was closed abnormally, e.g., without sending or receiving a Close control frame

In my terminal the following messages are visible:

$ ./marija 
Marija server started, listening on address 127.0.0.1:8080.
2016/11/30 13:46:47 Connection upgraded.
2016/11/30 13:46:53 Connection closed
2016/11/30 13:46:56 Connection upgraded.
2016/11/30 13:47:04 Connection closed
2016/11/30 13:47:07 Connection upgraded.

Any suggestions on what to do next?

ckuethe avatar Nov 30 '16 21:11 ckuethe

{
  "name" : "my-remote-server",
  "cluster_name" : "my-elk-stack",
  "cluster_uuid" : "KmBuokUjSnmsO7ZUYGygOA",
  "version" : {
    "number" : "5.0.2",
    "build_hash" : "f6b4951",
    "build_date" : "2016-11-24T10:07:18.101Z",
    "build_snapshot" : false,
    "lucene_version" : "6.2.1"
  },
  "tagline" : "You Know, for Search"
}

ckuethe avatar Nov 30 '16 22:11 ckuethe

@ckuethe did you use the latest commit? Or a release version?

nl5887 avatar Dec 01 '16 07:12 nl5887

Just pushed a version that disables sniffing (this resolves the ip from the Elasticsearch nodes), causing probably your issue.

nl5887 avatar Dec 01 '16 07:12 nl5887

Thanks for looking into this. Unfortunately even with the latest commit I still see this behavior... I'll clear my browser storage and do some more testing in the morning.

ckuethe avatar Dec 01 '16 08:12 ckuethe

I haven't tested it yet with ES 5 over a tunnel, but it works with older versions. Will setup some things to see if we can reproduce. Keep us posted, we're eager to solve this issue.

nl5887 avatar Dec 01 '16 11:12 nl5887

Something I noticed before, sometimes there are too many fields in the index (we'll look into this issue), could you remove all but the relevant indexes, and retry refreshing the fields?

nl5887 avatar Dec 01 '16 11:12 nl5887

There about 125 fields in each of the indices as I'm exploring syslog with a bunch of different hosts, log types, and fields courtesy of filebeat and logstash

ckuethe avatar Dec 01 '16 16:12 ckuethe

Could you export the mapping so we can try to reproduce the environment?

nl5887 avatar Dec 01 '16 19:12 nl5887

https://gist.github.com/ckuethe/a0c0ce3033c8250eb68e2362ffb30a85

ckuethe avatar Dec 01 '16 20:12 ckuethe

Oops - 142 fields.

screenshot_2016-12-01_12-33-06

ckuethe avatar Dec 01 '16 20:12 ckuethe

thx, we'll look into this issue

nl5887 avatar Dec 01 '16 22:12 nl5887