We need to provide individual per test log data. This is important for perf, because we're changing how we execute tests on swarming to execute many benchmarks in one swarming task, so it'll be much harder for sheriffs to debug broken tasks.

These should be links of some sort of sheriff-o-matic, which should show up per alert. There should also be a way to link to them externally.

I think a good solution for this is to store the logs on logdog, and come up with a consistent and deterministic naming convention. We then upload to these logdog streams from the telemetry code which executes, since it knows about the per story logging, or could learn very easily. SOM then also looks at this data and displays relevant links, if the logdog streams exist.

I think the android team does something somewhat similar to this with logcats, if I remember correctly? jbudorick@, how do android tests use logdog?