the fault domain setting of the daos container
|
dagouxiong2015@...
Hello everyone:
Recently, I am studying the fault domain setting of the daos container, hoping to deploy multiple engines on a single physical node,
and the data of the object is distributed on different physical nodes , prevent the failure of a single physical node to cause data loss.
But when I configure
[root@server-1 ~]# daos cont create pool --label cont12 --type POSIX --properties rf:1 --properties rf_lvl:0
ERROR: daos: "rf_lvl" is not a settable property (valid: cksum,cksum_size,compression,dedup,dedup_threshold,ec_cell,encryption,label,rf,srv_cksum,status)
Code analysis can only support one fault domain (by rank) layout. /**
* Level of fault-domain to use for object allocation
* rank is hardcoded to 1, [2-254] are defined by the admin
*/
enum {
DAOS_PROP_CO_REDUN_MIN = 1,
DAOS_PROP_CO_REDUN_RANK = 1, /** hard-coded */
DAOS_PROP_CO_REDUN_MAX = 254,
};
In the current situation, if you want data redundancy on different physical nodes, do you have any good suggestions?
Does daos plan to support configurable in fault domain in the future?
Best regards!
|
|
|
|
Johann
Hi there,
There is a gap in the support of rf_lvl which is not used by the placement algorithm yet. Please see https://daosio.atlassian.net/browse/DAOS-10215 to track progress on this.
Cheers, Johann
From:
<daos@daos.groups.io> on behalf of "dagouxiong2015@..." <dagouxiong2015@...>
Hello everyone: Recently, I am studying the fault domain setting of the daos container, hoping to deploy multiple engines on a single physical node, and the data of the object is distributed on different physical nodes , prevent the failure of a single physical node to cause data loss. But when I configure [root@server-1 ~]# daos cont create pool --label cont12 --type POSIX --properties rf:1 --properties rf_lvl:0 ERROR: daos: "rf_lvl" is not a settable property (valid: cksum,cksum_size,compression,dedup,dedup_threshold,ec_cell,encryption,label,rf,srv_cksum,status)
/** * Level of fault-domain to use for object allocation * rank is hardcoded to 1, [2-254] are defined by the admin */ enum { DAOS_PROP_CO_REDUN_MIN = 1, DAOS_PROP_CO_REDUN_RANK = 1, /** hard-coded */ DAOS_PROP_CO_REDUN_MAX = 254, };
In the current situation, if you want data redundancy on different physical nodes, do you have any good suggestions? Does daos plan to support configurable in fault domain in the future?
Best regards! --------------------------------------------------------------------- This e-mail and any attachments may contain confidential material for |
|
|