Log Reader

Overview

Plug-in Notes:

  • Allows only System Administrators or members of the Designers group to access log data.  
  • Does not load files outside the logs directory or files that do not end in .log.* or .csv.*.
  • On multi-application-server environments, logs will only be shown from a single application server.
  • Custom headers can be specified as an optional parameter. This is meant to be used for .csv files that don't contain a header.

Key Features & Functionality

The Appian Log Reader Plug-in contains functionality to:

  • Read files in the Appian /log folder (.csv or .log) and return them as structured data
  • Aggregate data from a .csv file
  • Serve as the source for a service-backed record for viewing logs
  • The Log Reader application provided with the download demonstrates a service-backed record for viewing logs, as well as reports on specific log files. Administrators can view reports on system health over time, including design best practices, system load, and database performance. The application also contains a process that checks details from system.csv and alerts administrators if memory or CPU utilization exceeds a threshold.
  • Tail a log file with tailcsv and taillog. Note that tail is an expensive operation which is optimized for reading the last few lines of a large log file. It's not advisable to tail the entire log file from the end to the beginning. However, tail will perform much better than the other log functions when reading the last few lines from a very large log file. Use the batch size and timestamp filters to limit the number of lines read by the tailcsv and taillog functions.
  • Takes a line of text in CSV format and returns a text array
Anonymous
Parents
  • Would it be possible to create a new version of the readCsvLog functions that doesn't strip out text-containing quotes as found in the original CSV file?  In particular the "Error Message" column is liable to have text that contains commas, and since each log row is returned as mere plaintext, in order to create a dictionary of data we're forced to use the split() function on commas.  But the function also removes the quote marks around strings that contain commas, so we have no way to verify we're splitting on the right things without really having to make big guesses.  I bring this up now because yet another corner case in the heuristics I was using to read the row data has cropped up, causing extra headaches.

    Honestly I'm not sure why this function doesn't return CDT or at least JSON data - that would make it incredibly less of a headache to use.

    As a concrete example, here's a row in the CSV file itself showing the error message wrapped in quotes:

    Whereas here's the same row, straight out of the readCsvLogPaging() function (notice the quotes have been stripped by the plug-in):

    CC: ,

  • Hey Mike is the owner of the plugin I will let him know about this. I did add a function called csvtotextarray. My expectation is that this function should correctly parse any CSV line returned by the smart service. I submitted the plugin update tonight.

Comment Children
  • At this point, this plug-in seems pretty much abandoned.  Any additional feedback here would be appreciated.

  • - are there any updates as to who's in charge of maintenance on this plug-in?  Is it being actively maintained / is there any chance of any of the issues or inconsistencies i've enumerated, being addressed in the foreseeable future?

  • Checking back in / , i can confirm that "csvToTextArray" fails to correctly parse CSV rows as returned by the original "readCsvLogPaging" function which (as i previously noted quite a while ago) incorrectly strips quote escaping from CSV text containing a comma (basically causing one cell worth of data to be treated as two cells when it contains a comma).

    As I noted somewhere, that seems to have been fixed in the "tailCsvLogPaging" function (i.e. passing a returned row from that function through "csvToTextArray" returns the expected number of fields, even when a row cell includes a comma); however there seem to still be unresolved issues even with that one - for one, the "headers" value seems to *only* return blank, and additionally, with no "start index" (as i noted a year or so ago) it leaves me unable to really execute my use case of creating a grid and allowing users to page through it (starting of course with most-recent-first).

    Any chance of some new changes to harden the behavior a bit and get them acting consistently, etc?  I have an Admin tool that queries error messages from the log but I have to bend over backwards to parse the rows carefully enough that it doesn't blow up on me, and it seems like every other week I need to install yet another heuristic to sanitize a comma appearing in a data row in a new way.  It's not really scalable.

  • that sounds good - for clarification does your new function assume the CSV text row will be "quote escaped" like it is in the original CSV file, or with quotes stripped like returned by the current "readCsvLog" functions?

    Also, do you mean the function has been added to this plug-in, whenever the update is published at least?