def start():
try:
settings=startup.read_settings()
Log.start(settings.debug)
main(settings)
except Exception, e:
Log.error("Problems exist", e)
def setUp(self):
settings = startup.read_settings(filename="leak_check_settings.json")
Log.start(settings.debug)
self.private = ElasticSearch(settings.private)
self.public = ElasticSearch(settings.public)
self.public_comments = ElasticSearch(settings.public_comments)
self.settings = settings
def start():
try:
settings = startup.read_settings(defs=[{
"name": ["--quick", "--fast"],
"help":
"use this to process the first and last block, useful for testing the config settings before doing a full run",
"action": "store_true",
"dest": "quick"
}, {
"name": ["--restart", "--reset", "--redo"],
"help":
"use this to force a reprocessing of all data",
"action":
"store_true",
"dest":
"restart"
}])
with startup.SingleInstance(flavor_id=settings.args.filename):
if settings.args.restart:
for l in struct.listwrap(settings.debug.log):
if l.filename:
File(l.filename).parent.delete()
File(settings.param.first_run_time).delete()
File(settings.param.last_run_time).delete()
Log.start(settings.debug)
main(settings)
except Exception, e:
Log.fatal("Can not start", e)
def test_private_etl(self):
"""
ENSURE IDENTIFIABLE INFORMATION DOES NOT EXIST ON ANY BUGS
"""
File(self.settings.param.first_run_time).delete()
File(self.settings.param.last_run_time).delete()
self.settings.param.allow_private_bugs = True
database.make_test_instance(self.settings.bugzilla)
es = elasticsearch.make_test_instance("candidate",
self.settings.fake.bugs)
es_comments = elasticsearch.make_test_instance(
"candidate_comments", self.settings.fake.comments)
bz_etl.main(self.settings, es, es_comments)
ref = elasticsearch.open_test_instance(
"reference", self.settings.private_bugs_reference)
compare_both(es, ref, self.settings, self.settings.param.bugs)
#DIRECT COMPARE THE FILE JSON
can = File(self.settings.fake.comments.filename).read()
ref = File(self.settings.private_comments_reference.filename).read()
if can != ref:
for i, c in enumerate(can):
found = -1
if can[i] != ref[i]:
found = i
break
Log.error("Comments do not match reference\n{{sample}}",
{"sample": can[MIN([0, found - 100]):found + 100]})
def get_pending(es, since):
result = es.search({
"query": {"filtered": {
"query": {"match_all": {}},
"filter": {
"range": {"modified_ts": {"gte": CNV.datetime2milli(since)}}}
}},
"from": 0,
"size": 0,
"sort": [],
"facets": {"default": {"terms": {"field": "bug_id", "size": 200000}}}
})
if len(result.facets.default.terms) >= 200000:
Log.error("Can not handle more than 200K bugs changed")
pending_bugs = Multiset(
result.facets.default.terms,
key_field="term",
count_field="count"
)
Log.note("Source has {{num}} bug versions for updating", {
"num": len(pending_bugs)
})
return pending_bugs
def start():
try:
settings = startup.read_settings()
Log.start(settings.debug)
main(settings, restart=True)
except Exception, e:
Log.error("Can not start", e)
def test_ambiguous_whiteboard_screened(self):
GOOD_BUG_TO_TEST=1046
database.make_test_instance(self.settings.bugzilla)
with DB(self.settings.bugzilla) as db:
es = elasticsearch.make_test_instance("candidate", self.settings.candidate)
#MARK BUG AS ONE OF THE SCREENED GROUPS
database.add_bug_group(db, GOOD_BUG_TO_TEST, SCREENED_WHITEBOARD_BUG_GROUPS[0])
#MARK BUG AS ONE OF THE *NOT* SCREENED GROUPS
database.add_bug_group(db, GOOD_BUG_TO_TEST, "not screened")
db.flush()
#SETUP RUN PARAMETERS
param = Struct()
param.end_time = CNV.datetime2milli(get_current_time(db))
param.start_time = 0
param.start_time_str = extract_bugzilla.milli2string(db, 0)
param.alias_file = self.settings.param.alias_file
param.bug_list = struct.wrap([GOOD_BUG_TO_TEST]) # bug 1046 sees lots of whiteboard, and other field, changes
param.allow_private_bugs = True
with ThreadedQueue(es, size=1000) as output:
etl(db, output, param, please_stop=None)
Thread.sleep(2) # MUST SLEEP WHILE ES DOES ITS INDEXING
versions = get_all_bug_versions(es, GOOD_BUG_TO_TEST)
for v in versions:
if v.status_whiteboard not in (None, "", "[screened]"):
Log.error("Expecting whiteboard to be screened")
def get_max_bug_id(es):
try:
results = es.search({
"query": {
"filtered": {
"query": {
"match_all": {}
},
"filter": {
"script": {
"script": "true"
}
}
}
},
"from": 0,
"size": 0,
"sort": [],
"facets": {
"0": {
"statistical": {
"field": "bug_id"
}
}
}
})
if results.facets["0"].count == 0:
return 0
return results.facets["0"].max
except Exception, e:
Log.error("Can not get_max_bug from {{host}}/{{index}}", {
"host": es.settings.host,
"index": es.settings.index
}, e)
def start():
try:
settings = startup.read_settings(defs=[{
"name": ["--quick", "--fast"],
"help": "use this to process the first and last block, useful for testing the config settings before doing a full run",
"action": "store_true",
"dest": "quick"
}, {
"name": ["--restart", "--reset", "--redo"],
"help": "use this to force a reprocessing of all data",
"action": "store_true",
"dest": "restart"
}])
with startup.SingleInstance(flavor_id=settings.args.filename):
if settings.args.restart:
for l in struct.listwrap(settings.debug.log):
if l.filename:
File(l.filename).parent.delete()
File(settings.param.first_run_time).delete()
File(settings.param.last_run_time).delete()
Log.start(settings.debug)
main(settings)
except Exception, e:
Log.fatal("Can not start", e)
def setup_es(settings, db, es, es_comments):
"""
SETUP ES CONNECTIONS TO REFLECT IF WE ARE RESUMING, INCREMENTAL, OR STARTING OVER
"""
current_run_time = get_current_time(db)
if File(settings.param.first_run_time).exists and File(settings.param.last_run_time).exists:
# INCREMENTAL UPDATE; DO NOT MAKE NEW INDEX
last_run_time = long(File(settings.param.last_run_time).read())
if not es:
es = ElasticSearch(settings.es)
es_comments = ElasticSearch(settings.es_comments)
elif File(settings.param.first_run_time).exists:
# DO NOT MAKE NEW INDEX, CONTINUE INITIAL FILL
try:
last_run_time = 0
current_run_time = long(File(settings.param.first_run_time).read())
if not es:
if not settings.es.alias:
temp = ElasticSearch(settings.es).get_proto(settings.es.index)
settings.es.alias = settings.es.index
settings.es.index = temp.last()
es = ElasticSearch(settings.es)
es.set_refresh_interval(1) #REQUIRED SO WE CAN SEE WHAT BUGS HAVE BEEN LOADED ALREADY
if not settings.es_comments.alias:
temp = ElasticSearch(settings.es_comments).get_proto(settings.es_comments.index)
settings.es_comments.alias = settings.es_comments.index
settings.es_comments.index = temp.last()
es_comments = ElasticSearch(settings.es_comments)
except Exception, e:
Log.warning("can not resume ETL, restarting", e)
File(settings.param.first_run_time).delete()
return setup_es(settings, db, es, es_comments)
def test_public_etl(self):
"""
ENSURE ETL GENERATES WHAT'S IN THE REFERENCE FILE
"""
File(self.settings.param.first_run_time).delete()
File(self.settings.param.last_run_time).delete()
self.settings.param.allow_private_bugs = Null
database.make_test_instance(self.settings.bugzilla)
es = elasticsearch.make_test_instance("candidate",
self.settings.fake.bugs)
es_comments = elasticsearch.make_test_instance(
"candidate_comments", self.settings.fake.comments)
bz_etl.main(self.settings, es, es_comments)
ref = elasticsearch.open_test_instance(
"reference", self.settings.public_bugs_reference)
compare_both(es, ref, self.settings, self.settings.param.bugs)
#DIRECT COMPARE THE FILE JSON
can = File(self.settings.fake.comments.filename).read()
ref = File(self.settings.public_comments_reference.filename).read()
if can != ref:
found = -1
for i, c in enumerate(can):
if can[i] != ref[i]:
found = i
break
Log.error("Comments do not match reference\n{{sample}}",
{"sample": can[MIN(0, found - 100):found + 100:]})
def setUp(self):
settings = startup.read_settings(filename="leak_check_settings.json")
Log.start(settings.debug)
self.private = ElasticSearch(settings.private)
self.public = ElasticSearch(settings.public)
self.public_comments = ElasticSearch(settings.public_comments)
self.settings = settings
def verify_public_bugs(es, private_bugs):
#VERIFY BUGS ARE IN OUTPUT
for b in private_bugs:
versions = compare_es.get_all_bug_versions(es, b)
if not versions:
Log.error("Expecting versions for public bug {{bug_id}}",
{"bug_id": b})
def test_whiteboard_screened(self):
GOOD_BUG_TO_TEST = 1046
database.make_test_instance(self.settings.bugzilla)
with DB(self.settings.bugzilla) as db:
es = elasticsearch.make_test_instance("candidate",
self.settings.candidate)
#MARK BUG AS ONE OF THE SCREENED GROUPS
database.add_bug_group(db, GOOD_BUG_TO_TEST,
SCREENED_WHITEBOARD_BUG_GROUPS[0])
db.flush()
#SETUP RUN PARAMETERS
param = Struct()
param.end_time = CNV.datetime2milli(get_current_time(db))
param.start_time = 0
param.start_time_str = extract_bugzilla.milli2string(db, 0)
param.alias_file = self.settings.param.alias_file
param.bug_list = struct.wrap([
GOOD_BUG_TO_TEST
]) # bug 1046 sees lots of whiteboard, and other field, changes
param.allow_private_bugs = True
with ThreadedQueue(es, size=1000) as output:
etl(db, output, param, please_stop=None)
Thread.sleep(2) # MUST SLEEP WHILE ES DOES ITS INDEXING
versions = get_all_bug_versions(es, GOOD_BUG_TO_TEST)
for v in versions:
if v.status_whiteboard not in (None, "", "[screened]"):
Log.error("Expecting whiteboard to be screened")
def get_recent_private_comments(db, param):
"""
GET COMMENTS THAT HAVE HAD THEIR PRIVACY INDICATOR CHANGED
"""
if param.allow_private_bugs:
return []
param.field_id = PRIVATE_COMMENTS_FIELD_ID
try:
comments = db.query(
"""
SELECT
a.comment_id,
a.bug_id
FROM
bugs_activity a
WHERE
bug_when >= {{start_time_str}} AND
fieldid={{field_id}}
""", param)
return comments
except Exception, e:
Log.error("problem getting recent private attachments", e)
def get_comments_by_id(db, comments, param):
"""
GET SPECIFIC COMMENTS
"""
if param.allow_private_bugs:
return []
param.comments_filter = esfilter2sqlwhere(db, {"and": [
{"term": {"isprivate": 0}},
{"terms": {"c.comment_id": comments}}
]})
try:
comments = db.query("""
SELECT
c.comment_id,
c.bug_id,
p.login_name modified_by,
UNIX_TIMESTAMP(CONVERT_TZ(bug_when, 'US/Pacific','UTC'))*1000 AS modified_ts,
c.thetext comment,
c.isprivate
FROM
longdescs c
LEFT JOIN
profiles p ON c.who = p.userid
LEFT JOIN
longdescs_tags t ON t.comment_id=c.comment_id AND t.tag <> 'deleted'
WHERE
{{comments_filter}}
""", param)
return comments
except Exception, e:
Log.error("can not get comment data", e)
def diff(db, table, old_record, new_record):
"""
UPDATE bugs_activity WITH THE CHANGES IN RECORDS
"""
now = milli2string(db, CNV.datetime2milli(get_current_time(db)))
changed = set(old_record.keys()) ^ set(new_record.keys())
changed |= set([k for k, v in old_record.items() if v != new_record[k]])
if table != u"bugs":
prefix = table + u"."
else:
prefix = u""
for c in changed:
fieldid=db.query("SELECT id FROM fielddefs WHERE name={{field_name}}", {"field_name": prefix + c})[0].id
if fieldid == None:
Log.error("Expecting a valid field name")
activity = Struct(
bug_id=old_record.bug_id,
who=1,
bug_when=now,
fieldid=fieldid,
removed=old_record[c],
added=new_record[c],
attach_id=old_record.attach_id,
comment_id=old_record.comment_id
)
db.insert("bugs_activity", activity)
db.execute("UPDATE bugs SET delta_ts={{now}} WHERE {{where}}", {
"now":now,
"where":esfilter2sqlwhere(db, {"term":{"bug_id":old_record.bug_id}})
})
def test_private_etl(self):
"""
ENSURE IDENTIFIABLE INFORMATION DOES NOT EXIST ON ANY BUGS
"""
File(self.settings.param.first_run_time).delete()
File(self.settings.param.last_run_time).delete()
self.settings.param.allow_private_bugs = True
database.make_test_instance(self.settings.bugzilla)
es = elasticsearch.make_test_instance("candidate", self.settings.fake.bugs)
es_comments = elasticsearch.make_test_instance("candidate_comments", self.settings.fake.comments)
bz_etl.main(self.settings, es, es_comments)
ref = elasticsearch.open_test_instance("reference", self.settings.private_bugs_reference)
compare_both(es, ref, self.settings, self.settings.param.bugs)
#DIRECT COMPARE THE FILE JSON
can = File(self.settings.fake.comments.filename).read()
ref = File(self.settings.private_comments_reference.filename).read()
if can != ref:
for i, c in enumerate(can):
found = -1
if can[i] != ref[i]:
found = i
break
Log.error("Comments do not match reference\n{{sample}}", {"sample": can[MIN([0, found - 100]):found + 100]})
def start():
try:
settings = startup.read_settings()
Log.start(settings.debug)
main(settings, restart=True)
except Exception, e:
Log.error("Can not start", e)
def test_public_etl(self):
"""
ENSURE ETL GENERATES WHAT'S IN THE REFERENCE FILE
"""
File(self.settings.param.first_run_time).delete()
File(self.settings.param.last_run_time).delete()
self.settings.param.allow_private_bugs = Null
database.make_test_instance(self.settings.bugzilla)
es = elasticsearch.make_test_instance("candidate", self.settings.fake.bugs)
es_comments = elasticsearch.make_test_instance("candidate_comments", self.settings.fake.comments)
bz_etl.main(self.settings, es, es_comments)
ref = elasticsearch.open_test_instance("reference", self.settings.public_bugs_reference)
compare_both(es, ref, self.settings, self.settings.param.bugs)
#DIRECT COMPARE THE FILE JSON
can = File(self.settings.fake.comments.filename).read()
ref = File(self.settings.public_comments_reference.filename).read()
if can != ref:
found = -1
for i, c in enumerate(can):
if can[i] != ref[i]:
found = i
break
Log.error("Comments do not match reference\n{{sample}}", {"sample": can[MIN(0, found - 100):found + 100:]})
def test_incremental_etl_catches_tracking_flags(self):
database.make_test_instance(self.settings.bugzilla)
with DB(self.settings.bugzilla) as db:
es = elasticsearch.make_test_instance("candidate", self.settings.candidate)
#SETUP RUN PARAMETERS
param = Struct()
param.end_time = CNV.datetime2milli(get_current_time(db))
# FLAGS ADDED TO BUG 813650 ON 18/12/2012 2:38:08 AM (PDT), SO START AT SOME LATER TIME
param.start_time = CNV.datetime2milli(CNV.string2datetime("02/01/2013 10:09:15", "%d/%m/%Y %H:%M:%S"))
param.start_time_str = extract_bugzilla.milli2string(db, param.start_time)
param.alias_file = self.settings.param.alias_file
param.bug_list = struct.wrap([813650])
param.allow_private_bugs = self.settings.param.allow_private_bugs
with ThreadedQueue(es, size=1000) as output:
etl(db, output, param, please_stop=None)
Thread.sleep(2) # MUST SLEEP WHILE ES DOES ITS INDEXING
versions = get_all_bug_versions(es, 813650)
flags = ["cf_status_firefox18", "cf_status_firefox19", "cf_status_firefox_esr17", "cf_status_b2g18"]
for v in versions:
if v.modified_ts>param.start_time:
for f in flags:
if v[f] != "fixed":
Log.error("813650 should have {{flag}}=='fixed'", {"flag": f})
def verify_no_private_bugs(es, private_bugs):
#VERIFY BUGS ARE NOT IN OUTPUT
for b in private_bugs:
versions = compare_es.get_all_bug_versions(es, b)
if versions:
Log.error("Expecting no version for private bug {{bug_id}}",
{"bug_id": b})
def tearDown(self):
#CLOSE THE CACHED DB CONNECTIONS
bz_etl.close_db_connections()
if all_db:
Log.error("not all db connections are closed")
Log.stop()
def tearDown(self):
#CLOSE THE CACHED DB CONNECTIONS
bz_etl.close_db_connections()
if all_db:
Log.error("not all db connections are closed")
Log.stop()
def verify_public_bugs(es, private_bugs):
#VERIFY BUGS ARE IN OUTPUT
for b in private_bugs:
versions = compare_es.get_all_bug_versions(es, b)
if not versions:
Log.error("Expecting versions for public bug {{bug_id}}", {
"bug_id": b
})
def verify_no_private_bugs(es, private_bugs):
#VERIFY BUGS ARE NOT IN OUTPUT
for b in private_bugs:
versions = compare_es.get_all_bug_versions(es, b)
if versions:
Log.error("Expecting no version for private bug {{bug_id}}", {
"bug_id": b
})
def verify_no_private_attachments(es, private_attachments):
#VERIFY ATTACHMENTS ARE NOT IN OUTPUT
for b in Q.select(private_attachments, "bug_id"):
versions = compare_es.get_all_bug_versions(es, b)
#WE ASSUME THE ATTACHMENT, IF IT EXISTS, WILL BE SOMEWHERE IN THE BUG IT
#BELONGS TO, IF AT ALL
for v in versions:
for a in v.attachments:
if a.attach_id in Q.select(private_attachments, "attach_id"):
Log.error("Private attachment should not exist")
def verify_no_private_attachments(es, private_attachments):
#VERIFY ATTACHMENTS ARE NOT IN OUTPUT
for b in Q.select(private_attachments, "bug_id"):
versions = compare_es.get_all_bug_versions(es, b)
#WE ASSUME THE ATTACHMENT, IF IT EXISTS, WILL BE SOMEWHERE IN THE BUG IT
#BELONGS TO, IF AT ALL
for v in versions:
for a in v.attachments:
if a.attach_id in Q.select(private_attachments, "attach_id"):
Log.error("Private attachment should not exist")
def get_private_bugs_for_delete(db, param):
if param.allow_private_bugs:
return {0} # NO BUGS TO DELETE
try:
with Timer("get all private bug ids"):
private_bugs = db.query("SELECT DISTINCT bug_id FROM bug_group_map")
return set(private_bugs.bug_id) | {0}
except Exception, e:
Log.error("problem getting private bugs", e)
def get_private_bugs_for_delete(db, param):
if param.allow_private_bugs:
return {0} # NO BUGS TO DELETE
try:
with Timer("get all private bug ids"):
private_bugs = db.query(
"SELECT DISTINCT bug_id FROM bug_group_map")
return set(private_bugs.bug_id) | {0}
except Exception, e:
Log.error("problem getting private bugs", e)
def run_both_etl(db, output_queue, es_comments, param):
comment_thread = Thread.run("etl comments", etl_comments, db, es_comments, param)
process_thread = Thread.run("etl", etl, db, output_queue, param)
result = comment_thread.join()
if result.exception:
Log.error("etl_comments had problems", result.exception)
result = process_thread.join()
if result.exception:
Log.error("etl had problems", result.exception)
def test_replication():
try:
settings=startup.read_settings(filename="replication_settings.json")
Log.start(settings.debug)
source=ElasticSearch(settings.source)
destination=replicate.get_or_create_index(settings["destination"], source)
replicate.replicate(source, destination, [537285], CNV.string2datetime("19900101", "%Y%m%d"))
finally:
Log.stop()
def extend(self, records):
"""
JUST SO WE MODEL A Queue
"""
records = {v["id"]: v["value"] for v in records}
struct.unwrap(self.data).update(records)
data_as_json = CNV.object2JSON(self.data, pretty=True)
File(self.filename).write(data_as_json)
Log.note("{{num}} items added", {"num": len(records)})
def run_both_etl(db, output_queue, es_comments, param):
comment_thread = Thread.run("etl comments", etl_comments, db, es_comments,
param)
process_thread = Thread.run("etl", etl, db, output_queue, param)
result = comment_thread.join()
if result.exception:
Log.error("etl_comments had problems", result.exception)
result = process_thread.join()
if result.exception:
Log.error("etl had problems", result.exception)
def main(settings, es=None, es_comments=None):
if not settings.param.allow_private_bugs and es and not es_comments:
Log.error("Must have ES for comments")
resume_from_last_run = File(
settings.param.first_run_time).exists and not File(
settings.param.last_run_time).exists
#MAKE HANDLES TO CONTAINERS
try:
with DB(settings.bugzilla, readonly=True) as db:
current_run_time, es, es_comments, last_run_time = setup_es(
settings, db, es, es_comments)
with ThreadedQueue(es, size=500, silent=True) as output_queue:
#SETUP RUN PARAMETERS
param = Struct()
param.end_time = CNV.datetime2milli(get_current_time(db))
# DB WRITES ARE DELAYED, RESULTING IN UNORDERED bug_when IN bugs_activity (AS IS ASSUMED FOR bugs(delats_ts))
# THIS JITTER IS USUALLY NO MORE THAN ONE SECOND, BUT WE WILL GO BACK 60sec, JUST IN CASE.
# THERE ARE OCCASIONAL WRITES THAT ARE IN GMT, BUT SINCE THEY LOOK LIKE THE FUTURE, WE CAPTURE THEM
param.start_time = last_run_time - nvl(
settings.param.look_back,
5 * 60 * 1000) # 5 MINUTE LOOK_BACK
param.start_time_str = extract_bugzilla.milli2string(
db, param.start_time)
param.alias_file = settings.param.alias_file
param.allow_private_bugs = settings.param.allow_private_bugs
if last_run_time > 0:
with Timer("run incremental etl"):
incremental_etl(settings, param, db, es, es_comments,
output_queue)
else:
with Timer("run full etl"):
full_etl(resume_from_last_run, settings, param, db, es,
es_comments, output_queue)
output_queue.add(Thread.STOP)
if settings.es.alias:
es.delete_all_but(settings.es.alias, settings.es.index)
es.add_alias(settings.es.alias)
if settings.es_comments.alias:
es.delete_all_but(settings.es_comments.alias,
settings.es_comments.index)
es_comments.add_alias(settings.es_comments.alias)
File(settings.param.last_run_time).write(
unicode(CNV.datetime2milli(current_run_time)))
except Exception, e:
Log.error("Problem with main ETL loop", e)
def open_test_instance(name, settings):
if settings.filename:
Log.note("Using {{filename}} as {{type}}", {
"filename": settings.filename,
"type": name
})
return Fake_ES(settings)
else:
Log.note("Using ES cluster at {{host}} as {{type}}", {
"host": settings.host,
"type": name
})
return ElasticSearch(settings)
def loadAliases(settings):
try:
try:
with Timer("load alias file at {{filename}}", {"filename":nvl(settings.param.alias_file.path, settings.param.alias_file)}):
alias_json = File(settings.param.alias_file).read()
except Exception, e:
Log.warning("No alias file found (looking at {{filename}}", {"filename":nvl(settings.param.alias_file.path, settings.param.alias_file)})
alias_json = "{}"
#self.aliases IS A dict POINTING TO structs
for k, v in CNV.JSON2object(alias_json).iteritems():
aliases[k] = struct.wrap(v)
Log.note("{{num}} aliases loaded", {"num": len(aliases.keys())})
def test_replication():
try:
settings = startup.read_settings(filename="replication_settings.json")
Log.start(settings.debug)
source = ElasticSearch(settings.source)
destination = replicate.get_or_create_index(settings["destination"],
source)
replicate.replicate(source, destination, [537285],
CNV.string2datetime("19900101", "%Y%m%d"))
finally:
Log.stop()
def get_comments(db, param):
if not param.bug_list:
return []
if param.allow_private_bugs:
param.comment_field = SQL("'[screened]' comment")
param.bug_filter = esfilter2sqlwhere(
db, {"and": [{
"terms": {
"bug_id": param.bug_list
}
}]})
else:
param.comment_field = SQL("c.thetext comment")
param.bug_filter = esfilter2sqlwhere(
db, {
"and": [{
"terms": {
"bug_id": param.bug_list
}
}, {
"term": {
"isprivate": 0
}
}]
})
try:
comments = db.query(
"""
SELECT
c.comment_id,
c.bug_id,
p.login_name modified_by,
UNIX_TIMESTAMP(CONVERT_TZ(bug_when, 'US/Pacific','UTC'))*1000 AS modified_ts,
{{comment_field}},
c.isprivate
FROM
longdescs c
LEFT JOIN
profiles p ON c.who = p.userid
LEFT JOIN
longdescs_tags t ON t.comment_id=c.comment_id AND t.tag <> 'deleted'
WHERE
{{bug_filter}} AND
bug_when >= {{start_time_str}}
""", param)
return comments
except Exception, e:
Log.error("can not get comment data", e)
def analysis(settings, last_run, please_stop):
DIFF = 7
if last_run:
DIFF = 4 #ONCE WE HAVE ALL THE DATA IN WE CAN BE LESS DISCRIMINATING
try_again = True
while try_again and not please_stop:
#FIND EMAIL MOST NEEDING REPLACEMENT
problem_agg = Multiset(allow_negative=True)
for bug_id, agg in bugs.iteritems():
#ONLY COUNT NEGATIVE EMAILS
for email, count in agg.dic.iteritems():
if count < 0:
problem_agg.add(alias(email), amount=count)
problems = Q.sort([
{"email": e, "count": c}
for e, c in problem_agg.dic.iteritems()
if not aliases.get(e, Null).ignore and (c <= -(DIFF / 2) or last_run)
], ["count", "email"])
try_again = False
for problem in problems:
if please_stop:
break
#FIND MOST LIKELY MATCH
solution_agg = Multiset(allow_negative=True)
for bug_id, agg in bugs.iteritems():
if agg.dic.get(problem.email, 0) < 0: #ONLY BUGS THAT ARE EXPERIENCING THIS problem
solution_agg += agg
solutions = Q.sort([{"email": e, "count": c} for e, c in solution_agg.dic.iteritems()], [{"field": "count", "sort": -1}, "email"])
if last_run and len(solutions) == 2 and solutions[0].count == -solutions[1].count:
#exact match
pass
elif len(solutions) <= 1 or (solutions[1].count + DIFF >= solutions[0].count):
#not distinctive enough
continue
best_solution = solutions[0]
Log.note("{{problem}} ({{score}}) -> {{solution}} {{matches}}", {
"problem": problem.email,
"score": problem.count,
"solution": best_solution.email,
"matches": CNV.object2JSON(Q.select(solutions, "count")[:10:])
})
try_again = True
add_alias(problem.email, best_solution.email)
saveAliases(settings)
def test_confidential_whiteboard_is_screened(self):
leaked_whiteboard = get(
self.private,
{
"and": [
{
"terms": {
"bug_group": SCREENED_WHITEBOARD_BUG_GROUPS
}
},
{
"exists": {
"field": "status_whiteboard"
}
},
{
"not": {
"terms": {
"status_whiteboard": ["", "[screened]"]
}
}
},
{
"range": {
"expires_on": {
"gte": NOW
}
}
}, #CURRENT RECORDS
{
"range": {
"modified_ts": {
"lt": A_WHILE_AGO
}
}
}, #OF A MINIMUM AGE
]
},
fields=[
"bug_id", "product", "component", "status_whiteboard",
"bug_group", "modified_ts"
],
limit=100)
if leaked_whiteboard:
for l in leaked_whiteboard:
l.modified_ts = CNV.datetime2string(
CNV.milli2datetime(l.modified_ts))
Log.error("Whiteboard leaking:\b{{leak}}",
{"leak": leaked_whiteboard})
def etl_comments(db, es, param, please_stop):
# CONNECTIONS ARE EXPENSIVE, CACHE HERE
with comment_db_cache_lock:
if not comment_db_cache:
comment_db = DB(db)
comment_db_cache.append(comment_db)
with comment_db_cache_lock:
Log.note("Read comments from database")
comments = get_comments(comment_db_cache[0], param)
for g, c in Q.groupby(comments, size=500):
with Timer("Write {{num}} comments to ElasticSearch", {"num": len(c)}):
es.extend({"id": cc.comment_id, "value": cc} for cc in c)
def etl_comments(db, es, param, please_stop):
# CONNECTIONS ARE EXPENSIVE, CACHE HERE
with comment_db_cache_lock:
if not comment_db_cache:
comment_db = DB(db)
comment_db_cache.append(comment_db)
with comment_db_cache_lock:
Log.note("Read comments from database")
comments = get_comments(comment_db_cache[0], param)
for g, c in Q.groupby(comments, size=500):
with Timer("Write {{num}} comments to ElasticSearch", {"num": len(c)}):
es.extend({"id": cc.comment_id, "value": cc} for cc in c)
def make_test_instance(db_settings):
if not db_settings.filename:
Log.note("Database schema will not be touched")
return
with Timer("Make database instance"):
try:
#CLEAR SCHEMA
Log.note("Make empty {{schema}} schema", {"schema":db_settings.schema})
no_schema=db_settings.copy()
no_schema.schema = None
with DB(no_schema) as db:
db.execute("DROP DATABASE IF EXISTS {{schema}}", {"schema":db.quote_column(db_settings.schema)})
db.execute("CREATE DATABASE {{schema}}", {"schema":db.quote_column(db_settings.schema)})
#FILL SCHEMA
Log.note("Fill {{schema}} schema with data", {"schema":db_settings.schema})
DB.execute_file(db_settings, db_settings.filename)
#ADD MISSING TABLES
with DB(db_settings) as db:
db.execute("""
CREATE TABLE `longdescs_tags` (
`id` mediumint(9) NOT NULL AUTO_INCREMENT,
`comment_id` int(11) DEFAULT NULL,
`tag` varchar(24) NOT NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `longdescs_tags_idx` (`comment_id`,`tag`),
CONSTRAINT `fk_longdescs_tags_comment_id_longdescs_comment_id` FOREIGN KEY (`comment_id`) REFERENCES `longdescs` (`comment_id`) ON DELETE CASCADE ON UPDATE CASCADE
) DEFAULT CHARSET=utf8""")
except Exception, e:
Log.error("Can not setup test database", e)
def test_incremental_has_correct_expires_on(self):
# 813650, 726635 BOTH HAVE CHANGES IN 2013
bugs = struct.wrap([813650, 726635])
start_incremental=CNV.datetime2milli(CNV.string2datetime("2013-01-01", "%Y-%m-%d"))
es = elasticsearch.make_test_instance("candidate", self.settings.candidate)
with DB(self.settings.bugzilla) as db:
#SETUP FIRST RUN PARAMETERS
param = Struct()
param.end_time = start_incremental
param.start_time = 0
param.start_time_str = extract_bugzilla.milli2string(db, param.start_time)
param.alias_file = self.settings.param.alias_file
param.bug_list = bugs
param.allow_private_bugs = False
with ThreadedQueue(es, size=1000) as output:
etl(db, output, param, please_stop=None)
#SETUP INCREMENTAL RUN PARAMETERS
param = Struct()
param.end_time = CNV.datetime2milli(datetime.utcnow())
param.start_time = start_incremental
param.start_time_str = extract_bugzilla.milli2string(db, param.start_time)
param.alias_file = self.settings.param.alias_file
param.bug_list = bugs
param.allow_private_bugs = False
with ThreadedQueue(es, size=1000) as output:
etl(db, output, param, please_stop=None)
for b in bugs:
results = es.search({
"query": {"filtered": {
"query": {"match_all": {}},
"filter": {"and":[
{"term":{"bug_id":b}},
{"range":{"expires_on":{"gte":CNV.datetime2milli(datetime.utcnow())}}}
]}
}},
"from": 0,
"size": 200000,
"sort": [],
"fields": ["bug_id"]
})
if results.hits.total>1:
Log.error("Expecting only one active bug_version record")
def verify_no_private_comments(es, private_comments):
data = es.search({
"query": {"filtered": {
"query": {"match_all": {}},
"filter": {"and": [
{"terms": {"comment_id": private_comments}}
]}
}},
"from": 0,
"size": 200000,
"sort": []
})
if Q.select(data.hits.hits, "_source"):
Log.error("Expecting no comments")
def test_private_comments_not_leaking(self):
leaked_comments = get(
self.public_comments,
{"term": {"isprivate": "1"}},
limit=20
)
if leaked_comments:
if self.settings.param.delete:
self.public_comments.delete_record(
{"terms":{"bug_id":leaked_comments.bug_id}}
)
Log.error("{{num}} comments marked private have leaked!\n{{comments|indent}}", {
"num": len(leaked_comments),
"comments": leaked_comments
})
def open_test_instance(name, settings):
if settings.filename:
Log.note("Using {{filename}} as {{type}}", {
"filename": settings.filename,
"type": name
})
return Fake_ES(settings)
else:
Log.note("Using ES cluster at {{host}} as {{type}}", {
"host": settings.host,
"type": name
})
ElasticSearch.delete_index(settings)
schema = CNV.JSON2object(File(settings.schema_file).read(), flexible=True, paths=True)
es = ElasticSearch.create_index(settings, schema, limit_replicas=True)
return es
def add_alias(lost, found):
found_record = aliases.get(found, None)
lost_record = aliases.get(lost, None)
new_canonical = found
old_canonical = nvl(lost_record.canonical, lost)
lost_record.canonical = new_canonical
delete_list = []
#FOLD bugs ON lost=found
for bug_id, agg in bugs.iteritems():
v = agg.dic.get(lost, 0)
if v != 0:
agg.add(lost, -v)
agg.add(found, v)
if not agg:
delete_list.append(bug_id)
#FOLD bugs ON old_canonical=new_canonical
if old_canonical != lost:
for bug_id, agg in bugs.iteritems():
v = agg.dic.get(old_canonical, 0)
if v != 0:
agg.add(old_canonical, -v)
agg.add(new_canonical, v)
if not agg:
delete_list.append(bug_id)
for d in delete_list:
del bugs[d]
#FOLD ALIASES
for k, v in aliases.iteritems():
if v.canonical == old_canonical:
Log.note(
"ALIAS REMAPPED: {{alias}}->{{old}} to {{alias}}->{{new}}", {
"alias": k,
"old": old_canonical,
"new": found
})
v.canonical = found
def extract_from_file(source_settings, destination):
with File(source_settings.filename) as handle:
for g, d in Q.groupby(handle, size=BATCH_SIZE):
try:
d2 = map(
lambda (x): {"id": x.id, "value": x},
map(
lambda(x): transform_bugzilla.normalize(CNV.JSON2object(fix_json(x))),
d
)
)
destination.add(d2)
except Exception, e:
filename = "Error_" + unicode(g) + ".txt"
File(filename).write(d)
Log.warning("Can not convert block {{block}} (file={{host}})", {
"block": g,
"filename": filename
}, e)
def get_or_create_index(destination_settings, source):
#CHECK IF INDEX, OR ALIAS, EXISTS
es = ElasticSearch(destination_settings)
aliases = es.get_aliases()
indexes = [a for a in aliases if a.alias == destination_settings.index]
if not indexes:
#CREATE INDEX
schema = source.get_schema()
assert schema.settings
assert schema.mappings
ElasticSearch.create_index(destination_settings, schema, limit_replicas=True)
elif len(indexes) > 1:
Log.error("do not know how to replicate to more than one index")
elif indexes[0].alias != None:
destination_settings.alias = destination_settings.index
destination_settings.index = indexes[0].index
return ElasticSearch(destination_settings)
def test_private_comments_not_leaking(self):
leaked_comments = get(self.public_comments,
{"term": {
"isprivate": "1"
}},
limit=20)
if leaked_comments:
if self.settings.param.delete:
self.public_comments.delete_record(
{"terms": {
"bug_id": leaked_comments.bug_id
}})
Log.error(
"{{num}} comments marked private have leaked!\n{{comments|indent}}",
{
"num": len(leaked_comments),
"comments": leaked_comments
})
def get_comments_by_id(db, comments, param):
"""
GET SPECIFIC COMMENTS
"""
if param.allow_private_bugs:
return []
param.comments_filter = esfilter2sqlwhere(db, {
"and": [{
"term": {
"isprivate": 0
}
}, {
"terms": {
"c.comment_id": comments
}
}]
})
try:
comments = db.query(
"""
SELECT
c.comment_id,
c.bug_id,
p.login_name modified_by,
UNIX_TIMESTAMP(CONVERT_TZ(bug_when, 'US/Pacific','UTC'))*1000 AS modified_ts,
c.thetext comment,
c.isprivate
FROM
longdescs c
LEFT JOIN
profiles p ON c.who = p.userid
LEFT JOIN
longdescs_tags t ON t.comment_id=c.comment_id AND t.tag <> 'deleted'
WHERE
{{comments_filter}}
""", param)
return comments
except Exception, e:
Log.error("can not get comment data", e)
def compare_both(candidate, reference, settings, some_bugs):
File(settings.param.errors).delete()
try_dir = settings.param.errors + "/try/"
ref_dir = settings.param.errors + "/ref/"
with Timer("Comparing to reference"):
found_errors = False
for bug_id in some_bugs:
try:
versions = Q.sort(
get_all_bug_versions(candidate, bug_id, datetime.utcnow()),
"modified_ts")
# WE CAN NOT EXPECT candidate TO BE UP TO DATE BECAUSE IT IS USING AN OLD IMAGE
if not versions:
max_time = CNV.milli2datetime(settings.bugzilla.expires_on)
else:
max_time = CNV.milli2datetime(versions.last().modified_ts)
pre_ref_versions = get_all_bug_versions(
reference, bug_id, max_time)
ref_versions = \
Q.sort(
#ADDED TO FIX OLD PRODUCTION BUG VERSIONS
[compare_es.old2new(x, settings.bugzilla.expires_on) for x in pre_ref_versions],
"modified_ts"
)
can = CNV.object2JSON(versions, pretty=True)
ref = CNV.object2JSON(ref_versions, pretty=True)
if can != ref:
found_errors = True
File(try_dir + unicode(bug_id) + ".txt").write(can)
File(ref_dir + unicode(bug_id) + ".txt").write(ref)
except Exception, e:
found_errors = True
Log.warning("Problem ETL'ing bug {{bug_id}}",
{"bug_id": bug_id}, e)
if found_errors:
Log.error("DIFFERENCES FOUND (Differences shown in {{path}})",
{"path": [try_dir, ref_dir]})
def main(settings):
file = File(settings.param.alias_file)
aliases = CNV.JSON2object(file.read())
for v in aliases.values():
v.candidates = CNV.dict2Multiset(v.candidates)
data = [{
"lost": n,
"found": d.canonical
} for n, d in aliases.items() if d.canonical != None and n != d.canonical]
sorted = Q.sort(data, "found")
for s in sorted:
Log.note("{{found}} == {{lost}}", s)
clean = {
n: d.canonical
for n, d in aliases.items()
if d.canonical != None and n != d.canonical and n != ""
}
rev_clean = struct.inverse(clean)
Log.note(CNV.object2JSON(rev_clean, pretty=True))
for k, v in rev_clean.items():
if len(v) > 3:
Log.note(CNV.object2JSON({k: v}, pretty=True))
