Hi,
While working at the transition from SIS to ODA I got the feeling that ODA is
over-designed. There are loads of fields that are not used (yet?), the tables
match the SIS tables poorly and key fields are used excessively.
Here's the trouble I'm facing:
SIS consists of three tables (with fields in following):
Adapter: client mac ip netmask devname
Client: route hostname domainname arch imagename name proccount
Image: arch name location
The corresponding data is designed to be stored in ODA in 4 tables:
Images: architecture id name path
Nics: id ip mac name network_id node_id
Nodes: cluster_id cpu_num cpu_speed dns_domain fqdn group_id hostname id
image_id installer name ram swap units virtual
Networks: base_ip broadcast cluster_id gateway high_ip n_id name netmask
rfc1918
The problems:
- mapping from SIS to ODA is hard, info which is found in one SIS table is
scattered around several ODA tables. While this can be solved for reading
data, (the list_* routines in NewDB.pm work), it leads to problems when
updating and deleting data.
- most of the data in ODA is not used and has been entered for an eventuality
which might come sometime in the future. Maybe it's easier to work with
something which is slimmer but easilly extendable.
- ODA uses "id" fields in a lot of places although other natural unique key
fields are around:
Images: "id" is used, although "name" is also a unique key
Nics: "id" is used but "ip" would be the natural unique key
Nodes: again: "id" versus "name" or "hostname"
Networks: "n_id" versus "name"
This is a waste of space and complicates the structure unnecessarily. Can we
please get rid of the duplicate key fields? Actually most of the "id" fields
could go away.
What am I supposed to do when deleting a node record? Delete the record,
delete its Nics and delete its Network if nobody else is using it? I guess I
should rather leave the Network alone...
When I create a node record, do I have to create the Network record, too, if
it doesn't exist already? As we don't have any means of dealing with networks
(except defining it for the master node), how was was the usage of that field
envisioned? When should it be created? Was a GUI or API planned for this?
Thanks for any comments on this!
Best regards,
Erich
-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel
