On the beta cluster, Parsoid runs on the instance deployment-parsoid05.eqiad.wmflabs. I noticed the four parsoid process eating all the CPU.
Looking at the log file at `/data/project/parsoid/parsoid.log` tgere are a bunch of socket timeout errors against eswiki/wikidatawiki db (at least):
```
lang=json
{
"name":"parsoid",
"hostname":"deployment-parsoid05",
"pid":22036,
"level":40,
"logType":"warning/api/esockettimedout",
"wiki":"wikidatawiki",
"title":"Q63760",
"oldId":null,
"reqId":"572eeab6459c44ffbf2e7a439858cfcd",
"msg":"Failed API request, {\"error\":{\"code\":\"ESOCKETTIMEDOUT\"},\"retries-remaining\":1}",
"longMsg":"Failed API request,\n{\"error\":{\"code\":\"ESOCKETTIMEDOUT\"},\"retries-remaining\":1}",
"time":"2015-04-28T15:27:21.925Z",
"v":0
}
```
https://logstash-beta.wmflabs.org/ gives a bit more informations using the filters:
```
field : _type == value : parsoid
logType == query : "warning/api/etimedout"
```
{F158161 size=full}