Race in write_concern_after_stepdown.js

We expect the parallel shell to begin a w: majority write on Node 0, while Node 0 is primary and replication is blocked. Then Node 0 steps down and the parallel shell expects to receive PrimarySteppedDown.

In fact, the parallel shell took a long time to begin. It finally connected after Node 0 had already stepped down. When it began its w: majority write it received

w:majority write replied: {

  "ok" : 0,

  "errmsg" : "not master",

  "code" : 10107,

  "codeName" : "NotMaster"

}

This is not a "write error" but a generic command failure, so when the shell calls assert.writeErrorWithCode it logs unhelpfully:

assert: found code(s) [ ] does not match any of the expecte d codes [ 189, 11602 ]