Replication of a virtual distributed volume

摘要

In one aspect, a method includes running a virtual machine (VM) at a first site, sending I/Os from the VM to a first virtual volume (VVol) at the first site, mirroring the first VVol with a second VVol at a second site, sending I/Os for the first VVol to a third VVol at a third site, detecting failure at the first site, running the VM at the second site, reading a second backlog, adding the second backlog to a delta marker on the second site, the delta marker storing differences between the third VVol and the second VVol and replicating the second VVol to the third VVol after adding the second backlog to the delta marker. The second backlog is synchronized with a first backlog on the first site and the first backlog stores I/Os that have not been completed to the third VVol.

主权项

1. A method comprising:
running a virtual machine (VM) at a first site; sending I/Os from the VM to a first virtual volume (VVol) at the first site; mirroring the first VVol with a second VVol at a second site; sending I/Os for the first VVol to a third VVol at a third site; detecting failure at the first site; running the VM at the second site; reading a second backlog, the second backlog being synchronized with a first backlog on the first site, the first backlog storing I/Os that have not been completed to the third VVol; adding the second backlog to a delta marker on the second site, the delta marker storing differences between the third VVol and the second VVol; replicating the second VVol to the third VVol after adding the second backlog to the delta marker; receiving a notification to select an erase point in the delta marker; receiving notification that latest I/O arrived at the third VVol; and erasing data in the delta marker up to the erase point.