Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, May 10, 2007 at 12:26:58PM -0700, Carson Gaspar wrote: > Lars Ellenberg wrote: > >On Thu, May 10, 2007 at 10:44:58AM +0100, Mark Watts wrote: > > >> B: write IO is reported as completed, if it has reached > >> local DISK and remote buffer cache. > > * for cases were you can live with possible data loss > > of the last few milliseconds in case of failover > > If the standby node has the data in its buffer cache, how is it lost in > a failover? Or does protocol B not wait for an ack from the secondary > before declaring the write successful (so you lose the bits "on the wire")? say, you have a common power failure. Primary completes transaction, while Secondary has it still in RAM only. application tells user: transaction completed. common power failure, both nodes go down now. after power up, for some reason the former Primary stays down. those last transactions were reported as completed to the user. but they are lost now... anyways, protocol B, though in theory it is a good idea, in practice it does not buy you much performance, in my experience. -- : Lars Ellenberg Tel +43-1-8178292-0 : : LINBIT Information Technologies GmbH Fax +43-1-8178292-82 : : Vivenotgasse 48, A-1120 Vienna/Europe http://www.linbit.com : __ please use the "List-Reply" function of your email client.