[Related but different than T233623: MediaWiki log spam during master unavailable / read only, this one is about spam during https://wikitech.wikimedia.org/wiki/Incident_documentation/20190920-d2_switch_failure]
We've observed a large amount of logs sent by MW during the incident above, namely a peak of ~130k msg/s (see below) with the biggest chuck being from memcache channel:
e.g. Memcached error for key "WANCache:t:ptwiki:messages:pt" on server "127.0.0.1:11213": SERVER ERROR (https://logstash.wikimedia.org/goto/8c177e65b10582aaf15a904b91977751)
while the rest was mostly dbconnection / dbreplication channels: https://logstash.wikimedia.org/goto/006c2fa6f31a1d6b7138e673cb077a20
There should be some logging rate limit (per mw channel + level?) so that similar situations are handled more gracefully