Pywikibot checks if maxlag is returned ( https://phabricator.wikimedia.org/diffusion/PWBC/browse/master/pywikibot/data/api.py;31c0234c3e0dd5f0740aa7d69d94651fbcb8bea8$2075 )
lagpattern = re.compile(r"Waiting for [\d.]+: (?P<lag>\d+) seconds? lagged")
if code == "maxlag": lag = lagpattern.search(info) if lag: pywikibot.log( u"Pausing due to database lag: " + info) self.site.throttle.lag(int(lag.group("lag"))) continue
Parsing a string in the api output is just horrible, but seems to be the only way to retrieve the maxlag output. The output was an integer in seconds, but now a float so the regex doesn't match anymore. Simple fix is to update the regex to r"Waiting for [\d.]+: (?P<lag>\d+(\.\d+)?) seconds? lagged"
Bigger question is if we want to know the maximum lag and if we want to parse strings. I see several options here:
- Just patch it up (short term solution, should do this whatever longer term we choose)
- Stop using the lag info for backing up and instead just use the standard incremental backing off
- Update the mediawiki api to include the lag in the output in a structured format so we don't have to do evil string parsing