Message ID | cover.1603347462.git.anand.jain@oracle.com (mailing list archive) |
---|---|
Headers | show |
Series | readmirror feature (read_policy sysfs and in-memory only approach) | expand |
On Thu, Oct 22, 2020 at 03:43:34PM +0800, Anand Jain wrote:
> v9: C coding style fixes in 1/3 and 3/3
So the point of adding the sysfs knobs is to allow testing various
mirror selection strategies, what exactly was discussed in the past. Do
you have patches for that as well? It does not need to be final and
polished but at least give us something to test.
On 10/22/20 3:43 AM, Anand Jain wrote: > v9: C coding style fixes in 1/3 and 3/3 > You can add Reviewed-by: Josef Bacik <josef@toxicpanda.com> to this series, lets get this in so we can focus on the actual changes to the policy. Thanks, Josef
On 24/10/20 1:04 am, David Sterba wrote: > On Thu, Oct 22, 2020 at 03:43:34PM +0800, Anand Jain wrote: >> v9: C coding style fixes in 1/3 and 3/3 > > So the point of adding the sysfs knobs is to allow testing various > mirror selection strategies, what exactly was discussed in the past. Do > you have patches for that as well? It does not need to be final and > polished but at least give us something to test. > Sure. I just sent out the patchset [1]. It provides read_policy: latency, device, and round-robin. [1] https://patchwork.kernel.org/project/linux-btrfs/list/?series=371023 Thanks, Anand
On Tue, Oct 27, 2020 at 09:52:11AM +0800, Anand Jain wrote: > > > On 24/10/20 1:04 am, David Sterba wrote: > > On Thu, Oct 22, 2020 at 03:43:34PM +0800, Anand Jain wrote: > >> v9: C coding style fixes in 1/3 and 3/3 > > > > So the point of adding the sysfs knobs is to allow testing various > > mirror selection strategies, what exactly was discussed in the past. Do > > you have patches for that as well? It does not need to be final and > > polished but at least give us something to test. > > > > Sure. I just sent out the patchset [1]. It provides read_policy: > latency, device, and round-robin. > > [1] > https://patchwork.kernel.org/project/linux-btrfs/list/?series=371023 Exporting more information from the devices would help us to decide but is there anything we can do just with what we have? Or eventually add our counters like for the in-flight requests or total bytes. The intention is not to duplicate what block layer does as we need to experiment with the stats and heuristics it's just for that purpose and we don't have to rely on other subsystem patches.
On 28/10/20 2:02 am, David Sterba wrote: > On Tue, Oct 27, 2020 at 09:52:11AM +0800, Anand Jain wrote: >> >> >> On 24/10/20 1:04 am, David Sterba wrote: >>> On Thu, Oct 22, 2020 at 03:43:34PM +0800, Anand Jain wrote: >>>> v9: C coding style fixes in 1/3 and 3/3 >>> >>> So the point of adding the sysfs knobs is to allow testing various >>> mirror selection strategies, what exactly was discussed in the past. Do >>> you have patches for that as well? It does not need to be final and >>> polished but at least give us something to test. >>> >> >> Sure. I just sent out the patchset [1]. It provides read_policy: >> latency, device, and round-robin. >> >> [1] >> https://patchwork.kernel.org/project/linux-btrfs/list/?series=371023 > > Exporting more information from the devices would help us to decide but > is there anything we can do just with what we have? Or eventually add > our counters like for the in-flight requests or total bytes. The > intention is not to duplicate what block layer does as we need to > experiment with the stats and heuristics it's just for that purpose and > we don't have to rely on other subsystem patches. > I could rewrote the latency patch without export of any new block layer functions. Patchset v1 will soon be sent to the ML. I hope that shall address your concern. Thanks, Anand