Details
-
Task
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Done
-
None
Description
Spider supports a high availability setting described in the following slides: https://www.slideshare.net/Kentoku/spider-ha-20100922dtt7
However, this feature is difficult to use properly, and it makes the Spider implementation quite complex. I (nayuta.yanagisawa) strongly recommend using other HA solutions for high availability and using Spider just for sharding or federation. So, we deprecate the feature.
Attachments
Issue Links
- blocks
-
-
- Open
-
Activity
| Field | Original Value | New Value |
|---|---|---|
| Fix Version/s | 10.11 [ 27614 ] |
| Issue Type | Bug [ 1 ] | Task [ 3 ] |
| Fix Version/s | 10.9 [ 26905 ] | |
| Fix Version/s | 10.11 [ 27614 ] |
| Fix Version/s | 10.8 [ 26121 ] |
| Fix Version/s | 10.7 [ 24805 ] |
| Priority | Major [ 3 ] | Critical [ 2 ] |
| Link | This issue blocks MDEV-28862 [ MDEV-28862 ] |
| Description |
Spider supports a high availability setting described in the following slides: https://www.slideshare.net/Kentoku/spider-ha-20100922dtt7
However, this feature is difficult to use properly, and it makes the Spider implementation quite complex. I (nayuta.yanagisawa) strongly recommend using other HA solutions for high availability and using Spider just for sharding or federation. So, we deprecate the feature. |
Spider supports a high availability setting described in the following slides: https://www.slideshare.net/Kentoku/spider-ha-20100922dtt7
However, this feature is difficult to use properly, and it makes the Spider implementation quite complex. I (nayuta.yanagisawa) strongly recommend using other HA solutions for high availability and using Spider just for sharding or federation. So, we deprecate the feature. * Deprecate all the plugin variables related to the feature. * Show warnings when multiple values are specified for a single table attribute. |
| Description |
Spider supports a high availability setting described in the following slides: https://www.slideshare.net/Kentoku/spider-ha-20100922dtt7
However, this feature is difficult to use properly, and it makes the Spider implementation quite complex. I (nayuta.yanagisawa) strongly recommend using other HA solutions for high availability and using Spider just for sharding or federation. So, we deprecate the feature. * Deprecate all the plugin variables related to the feature. * Show warnings when multiple values are specified for a single table attribute. |
Spider supports a high availability setting described in the following slides: https://www.slideshare.net/Kentoku/spider-ha-20100922dtt7
However, this feature is difficult to use properly, and it makes the Spider implementation quite complex. I (nayuta.yanagisawa) strongly recommend using other HA solutions for high availability and using Spider just for sharding or federation. So, we deprecate the feature. |
| Status | Open [ 1 ] | In Progress [ 3 ] |
| Assignee | Nayuta Yanagisawa [ JIRAUSER47117 ] | Alexey Botchkov [ holyfoot ] |
| Status | In Progress [ 3 ] | In Review [ 10002 ] |
| Assignee | Alexey Botchkov [ holyfoot ] | Nayuta Yanagisawa [ JIRAUSER47117 ] |
| Status | In Review [ 10002 ] | Stalled [ 10000 ] |
| Fix Version/s | 10.7.5 [ 27505 ] | |
| Fix Version/s | 10.8.4 [ 27503 ] | |
| Fix Version/s | 10.9.2 [ 27115 ] | |
| Fix Version/s | 10.7 [ 24805 ] | |
| Fix Version/s | 10.8 [ 26121 ] | |
| Fix Version/s | 10.9 [ 26905 ] | |
| Resolution | Done [ 10200 ] | |
| Status | Stalled [ 10000 ] | Closed [ 6 ] |
You should consider the case RocksDB not working with GaleraCluster
Aurelien_LEQUOY did you comment on the wrong ticket or could you expand on the comment in relation to the ticket?
I think Aurélien's comment is about other HA solutions that can be used instead of Spider's HA solution. Galera Cluster does not support RocksDB so that means the only safely fully automated HA solution that can be used in combination with Spider only supports InnoDB.
I mean, if you remove this there no anymore solution to get HA spider with RocksDB.
node spider ======> rocksdb (data node)
Because rocksdb not supported with GaleraCluster.
michaeldg, Aurelien_LEQUOY: thanks for explaining. How about using binlog with RocksDB for replication - is it not an option?
https://mariadb.com/kb/en/myrocks-and-replication/
says statement-based replication may break consistency, but one could use mixed or row-based replication, no?
it's can be an option like InnoDB, but complicated to manage and in case of crash you can be sure all will be fine, have to deal with delay, crash etc... (more simple to manage it with Spider).
For me I use spider with huge table, and it's non sense for me to keep it on InnoDB. (it's can be if you really want to be safe with your data).
I got an apps for a client who insert 1.3 billion of rows each days, and InnoDB is not usefull for this case. an other one who got a mariadb with 50 Tb (all data inserted in one month, and rolling partitions)
By the way i working on version 4 of Pmacontrol, this version will be able to manage and monitor spider.
and I am pretty sure there is some companies again who use Spider with TokuDB even if this engine not supported anymore.
Aurelien_LEQUOY: sorry I was not suggesting that you switch to innodb. I was asking whether you could use binlog instead of spider/galera for replication. Do you mean that binlog does not work with rocksdb? The kb entries I referenced in my comment above seem to indicate otherwise. Granted the kb could be unclear or outdated but it would be good to know whether this is the case.
Can Elkin chime in this discussion and tell us how well binlog supports rocksdb compared to its support of innodb?
holyfoot Please review: https://github.com/MariaDB/server/commit/e42ede2514216447fbd9b85fa2c46debd586bc30