On Wed, 2018-12-19 at 17:02 +0530, Muneendra Kumar M wrote: > Hi Martin, > In one of the patch "[PATCH 00/19] san_path_err & multipath ANA > support" > > you have mentioned that san_path_err_XXX has some merits > over marginal_path_err_XXX. > > Is this understanding correct if so could you please explain the > scenario > in which use case this was better. > > I can say Marginal_path_err_xx is superset of san_path_err_xx. If you think so, please explain how. Imagine a user who has configured san_path_err_threshold X san_path_err_forget_rate Y san_path_err_recovery_time Z Now this user is suppsed migrate to marginal_path settings. marginal_path_double_failed_time A marginal_path_err_sample_time B marginal_path_err_rate_threshold C marginal_path_err_recheck_gap_time D Can you provide a formula to calculate A,B,C,D such that the system behaves the same way (or "better") than previously with X, Y, Z? I have pondered this for a while and concluded that I can't. > If we need both san_path_err_xx , Marginal_path_err_xx then so many > configurations will really confuse the customers. True, the many different options are confusing. However, I don't think it becomes much worse by offering both methods. Both methods aren't easy to understand by themselves. Once users understand that these two parameter sets are mutually exclusive, I think they can deal with that. What we really need is easier set-up of either method (think of 2-3 sets of reasobable pre-set parameter values for different scenarios). I believe most admins are so intimidated by the complexity of the parameters and their interaction that they give up and use delay_xx_checks instead, or nothing at all. Unfortunately this is all based on guessing; we at least have no data if users are trying these parameters and if yes, what they are using. Martin -- Dr. Martin Wilck <mwilck@xxxxxxxx>, Tel. +49 (0)911 74053 2107 SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
