crashstats-tools

Command line tools and library for interacting with Crash Stats (https://crash-stats.mozilla.org/).

Code:	https://github.com/mozilla-services/crashstats-tools
Documentation:	Check the README.rst file
Changelog:	Check the HISTORY.rst file
Issue tracker:	https://github.com/mozilla-services/crashstats-tools/issues
License:	MPLv2
Chat:	#crashreporting matrix channel
Community Participation Guidelines:	https://github.com/mozilla-services/crashstats-tools/blob/main/CODE_OF_CONDUCT.md

Installing

crashstats-tools is available on PyPI. You can install it with pipx:

$ pipx install crashstats-tools

For developing crashstats-tools, clone the Git repository, create a virtual environment, and install crashstats-tools and dev dependencies with:

$ pip install -e '.[dev]'

Tools

supersearch

Usage: supersearch [OPTIONS]

  Performs a basic search on Crash Stats using the Super Search API and outputs
  the results.

  A basic search uses filters and can span multiple pages of results.

  If you want to use facets, aggregations, histograms, or cardinatlities, use
  supersearchfacet.

  There are two ways to supersearch:

  First, you can specify Super Search API fields to generate the query.

  For example:

  $ supersearch --product=Firefox --num=10

  Second, you can pass in a url from a Super Search on Crash Stats. This command
  will then pull out the filter parameters. You can override those parameters
  with command line arguments.

  $ supersearch --supersearch-url='https://crash-stats.mozilla.org/search/...'

  Make sure to use single quotes when specifying values so that your shell
  doesn't expand variables or parse escape sequences.

  By default, supersearch looks at crash report data for the last week. You can
  specify start and end date filters to change the window of crash reports you
  want to look at.

  For example:

  $ supersearch --date='>=2024-10-01' --date='<2024-10-15'

  You can specify returned fields using the Super Search field "_columns".

  For example:

  $ supersearch --_columns=uuid --_columns=product --_columns=build_id

  Results are tab-delimited by default. You can specify other output formats
  using "--format". Tabs and newlines in output are escaped.

  For list of available fields and Super Search API documentation, see:

  https://crash-stats.mozilla.org/documentation/supersearch/

  https://crash-stats.mozilla.org/documentation/supersearch/api/

  This requires an API token in order to search and get results for protected
  data fields. Using an API token also reduces rate-limiting. Set the
  CRASHSTATS_API_TOKEN environment variable to your API token value:

  CRASHSTATS_API_TOKEN=xyz supersearch ...

  To create an API token for Crash Stats, visit:

  https://crash-stats.mozilla.org/api/tokens/

  Remember to abide by the data access policy when using data from Crash Stats!
  The policy is specified here:

  https://crash-stats.mozilla.org/documentation/protected_data_access/

Options:
  --host TEXT                     host for system to fetch crashids from
                                  [default: https://crash-stats.mozilla.org]
  --supersearch-url TEXT          Super Search url to base query on
  --num TEXT                      number of crash ids you want or "all" for all
                                  of them  [default: 100]
  --headers / --no-headers        whether or not to show table headers
                                  [default: no-headers]
  --format [table|tab|csv|json|markdown]
                                  format to print output  [default: tab]
  --verbose / --no-verbose        whether to print debugging output  [default:
                                  no-verbose]
  --color / --no-color            whether or not to colorize output; note that
                                  color is shut off when stdout is not an
                                  interactive terminal automatically  [default:
                                  color]
  --dotenv / --no-dotenv          whether or not to load a .env file for
                                  environment variables  [default: no-dotenv]
  --help                          Show this message and exit.

Fetch 10 crash ids for Firefox:

$ supersearch --num=10 --product=Firefox

Fetch all crash reports that have libc2.30.so/E22A1E7AEF7C58504AF2C60A5AD3A7AE0 in the modules_in_stack field:

$ supersearch --modules_in_stack=libc2.30.so/E22A1E7AEF6C58504AF2C60A5AD3A7AE0

This is helpful when you need to reprocess crash reports after uploading symbols for a module that we didn't have symbols for.

Fetch all crash reports that have libgallium_dri.so in the modules_in_stack field:

$ supersearch --modules_in_stack='^libgallium_dri.so'

Fetch 57 crash ids that match a Super Search query:

$ supersearch --num=57 \
    --supersearch-url='https://crash-stats.mozilla.org/search/?release_channel=nightly&version=70.0a1&product=Firefox&_sort=-date'

Use single quotes for values so that your shell doesn't do any shell expansion.

Fetch uuid, product, version, and build_id for crash reports that have "OOM" in the signature:

$ supersearch --_columns=uuid --_columns=product --_columns=version \
    --_columns=build_id --signature='~OOM'

Note that this doesn't support Super Search aggregations--use the supersearchfacet command for that.

See Super Search API documentation for details on notation and fields:

• https://crash-stats.mozilla.org/documentation/supersearch/
• https://crash-stats.mozilla.org/documentation/supersearch/api/

supersearchfacet

Usage: supersearchfacet [OPTIONS]

  Fetches facet data from Crash Stats using Super Search

  There are two ways to run this:

  First, you can specify Super Search API fields to generate the query.

  For example:

  $ supersearchfacet --product=Firefox --_facets=version

  Second, you can pass in a url from a Super Search on Crash Stats. This command
  will then pull out the parameters. You can override those parameters with
  command line arguments.

  $ supersearchfacet --_facets=version \
      --supersearch-url='https://crash-stats.mozilla.org/search/...'

  Make sure to use single quotes when specifying values so that your shell
  doesn't expand variables.

  You can get a facet of a field using ``_facets``.

  For example, this filters on Firefox and returns a facet on version for the
  last 7 days (the default time range).

  $ supersearchfacet --product=Firefox --_facets=version

  You can get cardinality (number of possible values), too. For example, this
  shows the number of different versions for Firefox crash reports in the last 7
  days.

  $ supersearchfacet --product=Firefox --_facets=_cardinality.version

  You can perform histograms. For example, this shows you counts for products
  per day for the last week:

  $ supersearchfacet --_histogram.date=product --relative-range=1w

  You can get a cardinality for the data for a field. For example, this tells
  you how many build ids there were for Firefox 124:

  $ supersearchfacet --product=Firefox --version=124.0
  --_facets=_cardinality.build_id

  You can do nested aggregations. For example, this shows the count of crash
  reports by product by release channel:

  $ supersearchfacet --_aggs.product=release_channel

  This shows count of crash reports by product, version, cardinality of
  install_time:

  $ supersearchfacet --_aggs.product.version=_cardinality.install_time

  Make sure to specify at least one of ``_facets``, ``_aggs``, ``_histogram``,
  or ``_cardinality``.

  By default, returned data is in a table. Tabs and newlines in output is
  escaped. Use ``--format`` to specify a different output format.

  For list of available fields and Super Search API documentation, see:

  https://crash-stats.mozilla.org/documentation/supersearch/

  https://crash-stats.mozilla.org/documentation/supersearch/api/

  This requires an API token in order to search and get results for protected
  data fields. Using an API token also reduces rate-limiting. Set the
  CRASHSTATS_API_TOKEN environment variable to your API token value:

  CRASHSTATS_API_TOKEN=xyz supersearchfacet ...

  To create an API token for Crash Stats, visit:

  https://crash-stats.mozilla.org/api/tokens/

  Remember to abide by the data access policy when using data from Crash Stats!
  The policy is specified here:

  https://crash-stats.mozilla.org/documentation/protected_data_access/

Options:
  --host TEXT                     host for system to fetch facets from
                                  [default: https://crash-stats.mozilla.org]
  --supersearch-url TEXT          Super Search url to base query on
  --start-date TEXT               start date for range; 'YYYY-MM-DD' and 'YYYY-
                                  MM-DD HH:MM:SS' formats; defaults to 00:00:00
                                  when no time specified
  --end-date TEXT                 end date for range; 'YYYY-MM-DD' and 'YYYY-MM-
                                  DD HH:MM:SS' formats; defaults to 23:59:59
                                  when no time specified  [default: today]
  --relative-range TEXT           relative range ending on end-date  [default:
                                  7d]
  --format [table|tab|csv|markdown|json|raw]
                                  format to print output  [default: table]
  --verbose / --no-verbose        whether to print debugging output  [default:
                                  no-verbose]
  --color / --no-color            whether or not to colorize output; note that
                                  color is shut off when stdout is not an
                                  interactive terminal automatically  [default:
                                  color]
  --denote-weekends / --no-denote-weekends
                                  This will add a * for values that are
                                  datestamps and on a Saturday or Sunday.
                                  [default: no-denote-weekends]
  --leftover-count / --no-leftover-count
                                  Calculates the leftover that is the difference
                                  between the total minus the sum of all term
                                  counts  [default: no-leftover-count]
  --dotenv / --no-dotenv          whether or not to load a .env file for
                                  environment variables  [default: no-dotenv]
  --help                          Show this message and exit.

See the breakdown of crash reports by product for the last 7 days:

$ supersearchfacet --_facets=product

See crashes broken down by product and down by day for the last 7 days:

$ supersearchfacet --_histogram.date=product --relative=range=7d

Histograms, facets, aggs, and cardinality can be filtered using Super Search filters. See crashes in Firefox by release channel broken down by day for the last 7 days:

$ supersearchfacet \
    --_histogram.date=release_channel \
    --release_channel=nightly \
    --release_channel=beta \
    --release_channel=release \
    --release_channel=esr \
    --product=Firefox \
    --relative-range=7d

Get the table in Markdown for easy cut-and-paste into Markdown things:

$ supersearchfacet --_histogram.date=product --relative=range=7d \
    --format=markdown

See Super Search API documentation for details on notation and fields:

• https://crash-stats.mozilla.org/documentation/supersearch/
• https://crash-stats.mozilla.org/documentation/supersearch/api/

fetch-data

Usage: fetch-data [OPTIONS] OUTPUTDIR [CRASH_IDS]...

  Fetches crash data from Crash Stats (https://crash-stats.mozilla.org/) system.

  Given one or more crash ids via command line or stdin (one per line), fetches
  crash data and puts it in specified directory.

  Crash data is split up into directories: raw_crash/, dump_names/,
  processed_crash/, and directories with the same name as the dump type.

  https://antenna.readthedocs.io/en/latest/overview.html#aws-s3-file-hierarchy

  This requires an API token in order to download dumps and protected data.
  Using an API token also reduces rate-limiting. Set the CRASHSTATS_API_TOKEN
  environment variable to your API token value:

  CRASHSTATS_API_TOKEN=xyz fetch-data crashdata ...

  To create an API token for Crash Stats, visit:

  https://crash-stats.mozilla.org/api/tokens/

  Remember to abide by the data access policy when using data from Crash Stats!
  The policy is specified here:

  https://crash-stats.mozilla.org/documentation/protected_data_access/

Options:
  --host TEXT                   host to pull crash data from; this needs to
                                match CRASHSTATS_API_TOKEN value  [default:
                                https://crash-stats.mozilla.org]
  --overwrite / --no-overwrite  whether or not to overwrite existing data
                                [default: overwrite]
  --raw / --no-raw              whether or not to save raw crash data  [default:
                                raw]
  --dumps / --no-dumps          whether or not to save dumps  [default: no-
                                dumps]
  --processed / --no-processed  whether or not to save processed crash data
                                [default: no-processed]
  --workers INTEGER RANGE       how many workers to use to download data;
                                requires CRASHSTATS_API_TOKEN  [default: 1;
                                1<=x<=10]
  --stats / --no-stats          prints download stats for large fetch-data jobs;
                                if it's printing download stats, it's not
                                printing other things  [default: no-stats]
  --color / --no-color          whether or not to colorize output; note that
                                color is shut off when stdout is not an
                                interactive terminal automatically  [default:
                                color]
  --dotenv / --no-dotenv        whether or not to load a .env file for
                                environment variables  [default: no-dotenv]
  --help                        Show this message and exit.

This lets you download raw crash, dumps, and processed crash from Crash Stats.

Fetch processed crash data for specific crash id:

$ fetch-data --no-raw --no-dumps --processed 723cacd6-1684-420e-a1c7-f04240190731

Fetch raw crash data using supersearch command to generate crash ids:

$ supersearch --product=Firefox --num=10 | \
    fetch-data --raw --no-dumps --no-processed crashdir

reprocess

Usage: reprocess [OPTIONS] [CRASHIDS]...

  Sends specified crashes for reprocessing

  This requires CRASHSTATS_API_TOKEN to be set in the environment to a valid API
  token.

  To create an API token for Crash Stats, visit:

  https://crash-stats.mozilla.org/api/tokens/

  Note: If you're processing more than 10,000 crashes, you should use a sleep
  value that balances the rate of crash ids being added to the queue and the
  rate of crash ids being processed. For example, you could use "--sleep 10"
  which will sleep for 10 seconds between submitting groups of crashes.

  Also, if you're processing a lot of crashes, you should let us know before you
  do it.

Options:
  --host TEXT                     host for system to reprocess in  [default:
                                  https://crash-stats.mozilla.org]
  --sleep INTEGER                 how long in seconds to sleep before submitting
                                  the next group  [default: 1]
  --ruleset TEXT                  processor pipeline ruleset to use for
                                  reprocessing these crash ids
  --allow-many / --no-allow-many  don't prompt user about letting us know about
                                  reprocessing more than 10,000 crashes
                                  [default: no-allow-many]
  --color / --no-color            whether or not to colorize output; note that
                                  color is shut off when stdout is not an
                                  interactive terminal automatically  [default:
                                  color]
  --dotenv / --no-dotenv          whether or not to load a .env file for
                                  environment variables  [default: no-dotenv]
  --help                          Show this message and exit.

Reprocess an individual crash report:

$ reprocess 723cacd6-1684-420e-a1c7-f04240190731

Reprocess crash reports based on a supersearch:

$ supersearch --num=5 | reprocess

Note

The reprocess command requires that you set CRASHSTATS_API_TOKEN in your environment with an API token that has the "Reprocess Crashes" permission.

Note

If you intend to reprocess more than 10,000 crash reports, please tell us first.

API token

For supersearch and fetch-data, you need to use a API token to:

• download data containing personally identifiable information
• download security sensitive data
• get out from the shadow of extreme API use rate limiting

You need an API token for reprocess--it doesn't work without one.

If you have access, you can generate an API token here:

https://crash-stats.mozilla.org/api/tokens/

Once you have acquired one, set the CRASHSTATS_API_TOKEN environment variable when using crashstats-tools commands.

Remember to abide by the data access policy when using data from Crash Stats! The policy is specified here:

https://crash-stats.mozilla.org/documentation/memory_dump_access/

Use cases

These tools are helpful when downloading data for analysis as well as downloading data to test other tools with.

Example 1

I want to collect a bunch of crash report data to look at possible values of an annotation in Firefox crash reports that's not available in Super Search, yet.

Since I'm looking just at annotations, all I need is the raw crash.

I would do something like this:

$ mkdir crashdata
$ supersearch --product=Firefox --num=1000 | \
    fetch-data --raw --no-dumps --no-processed crashdata

Then I can use jq or whatever to look at the crash report data in crashdata/raw_crash/.

Example 2

I want to test out a new JIT analysis tool that works on minidump files.

I would write a script like this:

#!/bin/bash

CRASHSTATS_API_TOKEN=foo
DATADIR=./crashdata
CRASHIDS=$(supersearch --product=Firefox --num=1000)

mkdir -p "${DATADIR}"

for crashid in ${CRASHIDS}
do
    echo "crashid ${crashid}"
    fetch-data --raw --dumps --no-processed "${DATADIR}" "${crashid}"

    # Not all crash reports have dumps--we only want to run analysis
    # on the ones that do.
    if [[ -e "crashdata/dump/${crashid}" ]]
    then
        echo "analyze dump ${crashid}..."
        # run my tool on the dump
    fi
done

Example 3

I want to get a list of crash ids for today (2019-07-30) where DOMFissionEnabled exists in the crash report.

I would do this:

$ supersearch --date=">=2019-07-30" --date='<2019-07-31' \
    --dom_fission_enabled='!__null__'

Example 4

I want to see number of crash reports for the last 14 days broken down by day and by product where DOMFissionEnabled exists in the crash report.

I would do this:

$ supersearchfacet --period=daily --format=markdown --relative-range=14d \
    --dom_fission_enabled='!__null__' --_facets=product

Library

Further, this provides a library interface to some Crash Stats API endpoints:

crashstats_tools.libcrashstats

get_crash_annotations(crash_id, api_token=None, host=DEFAULT_HOST)

Fetches crash annotations for a given crash report.

If you don't provide an API token, then it only returns crash annotations that are marked public.

get_dump(crash_id, dump_name, api_token, host=DEFAULT_HOST)

Fetches dumps, memory reports, and other crash report binaries for given crash id.

This requires an api token.

get_processed_crash(crash_id, api_token=None, host=DEFAULT_HOST)

Fetches the processed crash for given crash id.

supersearch(params, num_results, host=DEFAULT_HOST, api_token=None, logger=None)

Performs a super search and returns generator of result hits.

This doesn't return facet, aggregation, cardinality, or histogram data. If you want that, use supersearch_facet.

supersearch_facet(params, api_token=None, host=DEFAULT_HOST, logger=None)

Performs a super search and returns facet data

Prior art and related projects

https://github.com/mozilla/libmozdata

Python library which has a Supersearch class for performing queries and a ProcessedCrash class for fetching processed crash data.

https://github.com/mozilla-services/socorro

Socorro (the code base for Crash Stats) has a Docker-based local dev environment which includes a series of commands for manipulating data.

https://socorro.readthedocs.io/en/latest/service/processor.html#processing-crashes