-
Notifications
You must be signed in to change notification settings - Fork 51
ciao: Single VM Developer Setup #125
Comments
@mcastelino Dual roles is already supported by SSNTP as far as I know. The roles are bit masks that can be or'ed together. The changes to launcher should be minimal and I think could be done before the scheduler changes are made. Nothing should break as long as you don't enable launcher in dual mode. I guess the simplest thing to do would be to add a new value for -network, e.g., dual. I can do this tomorrow. |
@sameo What is the impact to SSNTP roles. Currently we have two explicit roles when we create certificates. netagent and agent. So it may make sense to eliminate the netagent role and use agent for both roles. The ciao-launcher will send in the role via the bit mask anyway?
|
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: ciao-project#125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: ciao-project#125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
I'll add a developer option to scheduler to enable a non-default relaxed runtime mode where cn and nn would not be exclusive resource types. |
Added branch where this development will take place https://github.com/01org/ciao/tree/mcastelino/topic/singlemachine |
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: #125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
Initial commits from @sameo enable ssntp.Role as a bitmask. I've got a stack of additional ssntp.Role related patches and the scheduler change for combo AGENT|NETAGENT nodes sitting on master at https://github.com/tpepper/ciao. I'm not sure what the process is...should I just push my couple changes to the mcastelino/topic/singlemachine branch or will @mcastelino cherry pick them in? |
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: #125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
As per @mcastelino I've pushed a rebase plus my changes to the https://github.com/01org/ciao/tree/mcastelino/topic/singlemachine branch. |
@tpepper 2a71778 looks really good, thanks for that. I have some comments about 928d521:
Although 2 adds one ssntp call to the scheduler implementation, I think it's cleaner than 1. It's generic and can be used in many other places while 1 clutters the API a bit. |
ssntp.server.ClientRole() looks good to me. I'll rework in that direction. |
I think I did this wrong. I think what I want is in the session.srcRole not the server. |
@tpepper The peer role is session.destRole. See how we call ntf.DisconnectNotify() from server.go for example. |
Hmm. So srcRole is the server and destRole is the client? If that's always the case, can sssntp.session change s/srcRole/serverRole and s/destRole/clientRole? Either way some comments might be good there in the code. |
@tpepper Both client.go and server.go use sessions and srcRole is the client role in the former and the server role in the latter. |
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: #125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: #125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
Single VM with an mock keystone implementation is now fully functional.
|
The functionality is fully implemented in 092cc34 The code needs to be merged into master. As next steps
|
In dual network mode, ciao-launcher acts as both a launcher of normal VM instances and of CNCI instances. This is a partial implementation of: ciao-project#125 Signed-off-by: Mark Ryan <mark.d.ryan@intel.com>
All patches except scheduler have been merged into master with PR #196. Once the scheduler changes are pulled in and tested this issue can be closed. |
Moving documentation to wiki https://github.com/01org/ciao/wiki/HOWTO:-Single-VM-Development-Environment |
On Fri 03 Jun at 17:10:04 -0700 notifications@github.com said:
The scheduler functional changes are already in. The only thing not in is Tim Pepper -- Linux OS Engineering |
@tpepper Some changes from the branch (e.g. multi role handling) are not present in master. For example This multi role handling has to be added to master. |
On Mon 06 Jun at 08:22:59 -0700 notifications@github.com said:
In master it's up in the one-higher-level function: https://github.com/01org/ciao/blob/master/ciao-scheduler/scheduler.go#L340 There the disconnect helper is called for each type that matches. Tim Pepper -- Linux OS Engineering |
Fully implemented and integrated with 89c764f |
Issues #124 documents how to setup a ciao developer setup on a bare metal system. However this setup is not optimal in some cases.
In this case the ciao should be configured to a special all in one mode where a node has dual role (i.e launcher can be a NN and a CN)
High Level Setup
Host Node (VM)
A clear VM running running on a NATed Virtual network with its own DHCP server.
This NATed Virtual Network network configuration is available by default using virt-manager or virsh on most linux distributions.
So the host is also the CN, NN, Controller , WebUI and Scheduler (and other openstack services) all run native on the host.
Setup
When the system is functioning the overall setup manifest as follows
As you can see below the CNCI VM's will end up on the NATed network. If this setup is run on a bare metal system the CNCI VM's will end up on the physical network. The Tenant VM's are invisible to host network or the NATed network The Bridge on which the NATed network is created should support hairpin mode and be configured as such.
Changes Needed to ciao components
In addition to the changes needed in #124
The text was updated successfully, but these errors were encountered: