[sheepdog] [PATCH 2/2] sheep: remove stale objects after recovery fully completed

MORITA Kazutaka morita.kazutaka at lab.ntt.co.jp
Mon Aug 27 06:06:57 CEST 2012


At Mon, 27 Aug 2012 11:58:26 +0800,
Liu Yuan wrote:
> 
> On 08/26/2012 03:54 PM, MORITA Kazutaka wrote:
> > +static void notify_recovery_completion_work(struct work *work)
> > +{
> > +	struct recovery_work *rw = container_of(work, struct recovery_work,
> > +						work);
> > +	struct sd_req hdr;
> > +	int ret;
> > +
> > +	sd_init_req(&hdr, SD_OP_COMPLETE_RECOVERY);
> > +	hdr.epoch = rw->epoch;
> > +	hdr.flags = SD_FLAG_CMD_WRITE;
> > +	hdr.data_length = sizeof(sys->this_node);
> > +
> > +	ret = exec_local_req(&hdr, &sys->this_node);
> 
> Why not send request directly rather than send it via gateway?

I just copied notify_vdi_deletion code.

What do you mean by "send request directly"?  Does it mean call
queue_request() directly, or call exec_req() against one of sheep
nodes?

Thanks,

Kazutaka



More information about the sheepdog mailing list