Repair checkpoint (.ckpt) files for various Splunk TAs

fix_checkpoints.py

"""
Helps to recover corrupted .ckpt files created by Splunks 'splunksdc' module used by various TA's.
(I think it ships with the TA add onbuilder, but not 100% sure of that.)

Requires the 'umsgpack' 3rd party module

Usage:

    python fix_checkpoint.py


More realistic usage:

    pip install umsgpack
    find $SPLUNK_HOME/var/lib/splunk/modinputs/ -name '*.ckpt' | xargs python fix_checkpoint.py
    
    # Rename any reparied files by hand:  (New file with .rebuild is created for any broken files)
    cd $SPLUNK_HOME/var/lib/splunk/modinputs/<inputtype>/
    mv MyInput.ckpt MyInput.ckpt.broken
    mv MyInput.ckpt.rebuild MyInput.ckpt


Helps with exceptions like this:

    2018-10-11 19:51:29,018 level=ERROR pid=30260 tid=MainThread logger=splunk_ta_o365.modinputs.management_activity pos=utils.py:wrapper:67 | datainput="Management_Activity_AzureActiveDirectory" start_time=1539287488 | message="Data input was interrupted by an unhandled exception." 
    Traceback (most recent call last):
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunksdc/utils.py", line 65, in wrapper
        return func(*args, **kwargs)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunk_ta_o365/modinputs/management_activity.py", line 88, in run
        with app.open_checkpoint(self.name) as checkpoint:
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunksdc/collector.py", line 258, in open_checkpoint
        checkpoint = LocalKVStore.open_always(fullname)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunksdc/checkpoint.py", line 167, in open_always
        indexes = cls.build_indexes(fp)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunksdc/checkpoint.py", line 174, in build_indexes
        for flag, key, pos in cls._replay(fp):
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/splunksdc/checkpoint.py", line 103, in _replay
        flag, key, _ = umsgpack.unpack(fp)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/3rdparty/umsgpack.py", line 879, in _unpack2
        return _unpack(fp, options)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/3rdparty/umsgpack.py", line 833, in _unpack
        return _unpack_dispatch_table[code](code, fp, options)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/3rdparty/umsgpack.py", line 785, in _unpack_array
        return [_unpack(fp, options) for i in xrange(length)]
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/3rdparty/umsgpack.py", line 833, in _unpack
        return _unpack_dispatch_table[code](code, fp, options)
      File "/opt/splunk/etc/apps/splunk_ta_o365/bin/3rdparty/umsgpack.py", line 699, in _unpack_string
        raise InvalidStringException("unpacked string is invalid utf-8")
    InvalidStringException: unpacked string is invalid utf-8


Lowell Alleman <lowell@kintyre.co>
Oct 11, 2018

Provided "AS IS".  No warranties.  This may be unfit for and or all purposes.
USE AT YOUR OWN RISK.  Things may melt or explode.   

"""
import sys
import umsgpack

def parse_msg(obj):
    # Really really basic test to ensure that the data is a triple to filter out garbage
    # that could have resulted from a corrupted record.  Not perect, but this was
    # 'good-enough' for my initial use case.
    action, _, _ = obj
    assert action in (0,1)

def test(checkpoint_file):
    with open(checkpoint_file, "r+b") as fp:
        # Skip magic block
        magic = fp.read(4)
        assert magic == "BUK0"
        c = 0
        for obj in umsgpack.Unpacker(fp, raw=False):
            parse_msg(obj)
            c += 1
            # print obj
    return c

def rebuild(checkpoint_file, checkpoint_new):
    print("Recovering {} into new file {}".format(checkpoint_file, checkpoint_new))
    with open(checkpoint_file, "r+b") as fp, open(checkpoint_new,"w+b") as np:
        # Skip magic block
        magic = fp.read(4)
        assert magic == "BUK0"
        np.write(magic)
        good = 0
        bad = 0
        iterable = umsgpack.Unpacker(fp, raw=False)
        while True:
            try:
                obj = next(iterable)
                print(obj)
                parse_msg(obj)
                umsgpack.pack(obj, np)
                good += 1
            except StopIteration:
                break
            except Exception as e:
                print("Exception: {}".format(e))
                bad += 1
    print("Recovered {} record, {} failed.  Saved into {}".format(good, bad, checkpoint_new))
    return good, bad


if __name__ == '__main__':

    for filename in sys.argv[1:]:
        try:
            lines = test(filename)
            print("Read {} with {} entries".format(filename, lines))
        except (TypeError, umsgpack.UnpackValueError) as e:
            print("Failed to process {}  {}".format(filename, e))
            rebuild(filename, filename + ".rebuild")

Additional background here: https://answers.splunk.com/answers/687115/what-should-i-do-about-the-following-office-365-ad.html?childToView=692238#answer-692238