In cluster/engine.go, we compute a delta between the manager and the engine. However, we do several blocking operations between the time we get the engine info, and the time we get the manager system. If those blocking operations take an abnormally long time, the user will be warned of a long time split, and the delta will be incorrect.
This patch moves computation of the delta to immediately after the Info call, so that the subsequent blocking operations do not affect it.
nishanttotla
commented
6 years ago
In cluster/engine.go, we compute a delta between the manager and the engine. However, we do several blocking operations between the time we get the engine info, and the time we get the manager system. If those blocking operations take an abnormally long time, the user will be warned of a long time split, and the delta will be incorrect.
This patch moves computation of the delta to immediately after the Info call, so that the subsequent blocking operations do not affect it.