[ML] Painless code in Watcher syntax created by ML throws IOException

[elasticmachine]

Original comment by @richcollier:

Customer discovered the following error upon watch execution for the code that the ML builds for the user automatically when the user creates a watch from the ML UI:

unchecked_i_o_exception: java.io.IOException: can not write type [class java.time.LocalDateTime]

The error originates from the following scripts within the watch:

                    "script_fields": {
                      "start": {
                        "script": {
                          "lang": "painless",
                          "inline": "LocalDateTime.ofEpochSecond((doc[\"timestamp\"].date.getMillis()-((doc[\"bucket_span\"].value * 1000)\n * params.padding)) / 1000, 0, ZoneOffset.UTC)",
                          "params": {
                            "padding": 10
                          }
                        }
                      },
                      "end": {
                        "script": {
                          "lang": "painless",
                          "inline": "LocalDateTime.ofEpochSecond((doc[\"timestamp\"].date.getMillis()+((doc[\"bucket_span\"].value * 1000)\n * params.padding)) / 1000, 0, ZoneOffset.UTC)",
                          "params": {
                            "padding": 10
                          }
                        }
                      },

To make it work, we need to add the .toString() to cast the LocalDateTime object into a String.:

                    "script_fields": {
                      "start": {
                        "script": {
                          "lang": "painless",
                          "inline": "LocalDateTime.ofEpochSecond((doc[\"timestamp\"].date.getMillis()-((doc[\"bucket_span\"].value * 1000)\n * params.padding)) / 1000, 0, ZoneOffset.UTC).toString()",
                          "params": {
                            "padding": 10
                          }
                        }
                      },
                      "end": {
                        "script": {
                          "lang": "painless",
                          "inline": "LocalDateTime.ofEpochSecond((doc[\"timestamp\"].date.getMillis()+((doc[\"bucket_span\"].value * 1000)\n * params.padding)) / 1000, 0, ZoneOffset.UTC).toString()",
                          "params": {
                            "padding": 10
                          }
                        }
                      },

I think this might only show up in certain situations (multi-node clusters?) since I couldn't replicate this on my 1-node setup.

The customer has confirmed that the above modification avoids the error and thus it should be incorporated into the build.

[jeffrey-e]

Hi, just confirmed this solution as I was searching for this issue. I came across the issue after upgrading my All in 1 machine to 6.3.1. Hopefully this gets fixed in the next version as it seems as a very minor change