Example: https://logstash.wikimedia.org/app/kibana#/doc/logstash-*/logstash-2020.11.04/clienterror/?id=AXWU8DBWG2-KaiAts1s7

I blame https://de.wikipedia.org/w/index.php?title=A&banner=B20_WMDE_Test_09_ctrl&uselang=de&force=1 - the other banner that's running is https://de.wikipedia.org/w/index.php?title=A&banner=wikipedia_asian_month_2020&uselang=de&force=1 and that seems to be less complicated.

Timing coincides with campaign C20_WMDE_Test_09 going up, so yes it's probably one or both of B20_WMDE_Test_09_ctrl or B20_WMDE_Test_09_var. Pinging WMDE-Fundraising-Tech as owners of those banners.

WMF Fundraising saw a similar issue before in T264366#6510591 caused by corrupt Firefox profiles (https://stackoverflow.com/questions/18877643/error-in-local-storage-ns-error-file-corrupted-firefox/26371494), and all of these errors also appear to be in Firefox. We fixed it by wrapping the use of localStorage in a try/catch. Alternatively I believe the mw.storage module can handle this.

Bump.
This has been UBN for almost 2 weeks now. Is there some one we can directly ping @Pcoombe to ensure this gets attention?

@Jdlrobson I'll have a look at this issue.

Thank you @kai.nissen !

Seems that @Pcoombe's comment regarding corrupt Firefox profiles points into the right direction.

I could not reproduce this using any specific browser functionality, but can confirm that this only happens in Firefox. This error has been reported to logstash since Sep 23rd and at a much lower rate during earlier banner tests, probably because of the traffic limit. There is a spike in early October which coincides with the Wikipedia Challenge banners.

Note: currently we are not capturing all errors for Chrome due to a bug in the instrumentation (T266517), so it is very likely the numbers are much higher and do include Chrome.

I've created a patch for our localstorage access, to guard against corruption in Firefox: https://github.com/wmde/fundraising-banners/pull/466

After redeploying all banners, the occurrences dropped to 0 (see Logstash).