Skip to content
This repository has been archived by the owner on Jun 6, 2024. It is now read-only.

Commit

Permalink
Change gpuType(s) to skuType(s)
Browse files Browse the repository at this point in the history 
Change gpuType(s) to skuType(s).
  • Loading branch information
@abuccts
abuccts committed Apr 3, 2020
1 parent 9838419 commit 279c0e5
Show file tree
Hide file tree
Showing 3 changed files with 26 additions and 20 deletions.
2 changes: 1 addition & 1 deletion src/rest-server/src/config/v2/hived.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ const hivedSchema = {
'^[A-Za-z0-9._~]+$': {
type: 'object',
properties: {
gpuType: {
skuType: {
type: ['string', 'null'],
default: null,
},
Expand Down
10 changes: 8 additions & 2 deletions src/rest-server/src/config/vc.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -48,8 +48,14 @@ const vcStatusPutInputSchema = Joi.object().keys({
const resourceUnits = {};

if (enabledHived) {
const hivedSpec = yaml.safeLoad(fs.readFileSync(hivedSpecPath));
for (let [key, val] of Object.entries(hivedSpec.physicalCluster.gpuTypes)) {
const hivedConfig = yaml.safeLoad(fs.readFileSync(hivedSpecPath));
if (!(physicalCluster in hivedConfig &&
skuTypes in hivedConfig.physicalCluster &&
Object.keys(hivedConfig.physicalCluster.skuTypes).length > 0)) {
new Error('Cannot find skuTypes in hivedscheduler config.');
}

for (let [key, val] of Object.entries(hivedConfig.physicalCluster.skuTypes)) {
resourceUnits[key] = {
cpu: k8s.atoi(val.cpu),
memory: k8s.convertMemoryMb(val.memory),
Expand Down
34 changes: 17 additions & 17 deletions src/rest-server/src/middlewares/v2/hived.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -90,50 +90,50 @@ const hivedValidate = async (protocolObj, username) => {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: ${taskRole} does not exist.`
`Taskrole ${taskRole} does not exist.`
);
}

const taskRoleConfig = hivedConfig.taskRoles[taskRole];
// at most one of [reservationId, gpuType] allowed
if (taskRoleConfig.reservationId !== null && taskRoleConfig.gpuType !== null) {
// at most one of [reservationId, skuType] allowed
if (taskRoleConfig.reservationId !== null && taskRoleConfig.skuType !== null) {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: ${taskRole} has both reservationId and gpuType, only one allowed.`
`Taskrole ${taskRole} has both reservationId and skuType, only one allowed.`
);
}

if (taskRoleConfig.gpuType !== null && !(taskRoleConfig.gpuType in resourceUnits)) {
if (taskRoleConfig.skuType !== null && !(taskRoleConfig.skuType in resourceUnits)) {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: ${taskRole} has unknown gpuType ${taskRoleConfig.gpuType}, allow ${Object.keys(resourceUnits)}.`
`Taskrole ${taskRole} has unknown skuType ${taskRoleConfig.skuType}, allow ${Object.keys(resourceUnits)}.`
);
}

const affinityGroupName = taskRoleConfig.affinityGroupName;
// affinityGroup should have uniform reservationId and gpuType
// affinityGroup should have uniform reservationId and skuType
if (affinityGroupName !== null) {
if (affinityGroupName in affinityGroups) {
if (taskRoleConfig.reservationId === null) {
taskRoleConfig.reservationId = affinityGroups[affinityGroupName].reservationId;
}
if (taskRoleConfig.gpuType === null) {
taskRoleConfig.gpuType = affinityGroups[affinityGroupName].gpuType;
if (taskRoleConfig.skuType === null) {
taskRoleConfig.skuType = affinityGroups[affinityGroupName].skuType;
}
if (taskRoleConfig.reservationId !== affinityGroups[affinityGroupName].reservationId ||
taskRoleConfig.gpuType !== affinityGroups[affinityGroupName].gpuType) {
taskRoleConfig.skuType !== affinityGroups[affinityGroupName].skuType) {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: affinityGroup: ${affinityGroupName} has inconsistent gpuType or reservationId.`
`AffinityGroup ${affinityGroupName} has inconsistent skuType or reservationId.`
);
}
} else {
affinityGroups[affinityGroupName] = {
reservationId: taskRoleConfig.reservationId,
gpuType: taskRoleConfig.gpuType,
skuType: taskRoleConfig.skuType,
affinityTaskList: [],
};
}
Expand All @@ -146,12 +146,12 @@ const hivedValidate = async (protocolObj, username) => {
}

for (let affinityGroupName of Object.keys(affinityGroups)) {
if (affinityGroups[affinityGroupName].gpuType !== null &&
if (affinityGroups[affinityGroupName].skuType !== null &&
affinityGroups[affinityGroupName].reservationId !== null) {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: affinityGroup: ${affinityGroupName} has both reservationId and gpuType, only one allowed.`
`AffinityGroup ${affinityGroupName} has both reservationId and skuType, only one allowed.`
);
}
}
Expand Down Expand Up @@ -194,7 +194,7 @@ const hivedValidate = async (protocolObj, username) => {
affinityGroup: null,
};
if (hivedConfig && hivedConfig.taskRoles && taskRole in hivedConfig.taskRoles) {
podSpec.gpuType = hivedConfig.taskRoles[taskRole].gpuType;
podSpec.gpuType = hivedConfig.taskRoles[taskRole].skuType;
if (podSpec.gpuType !== null) {
for (const t of ['gpu', 'cpu', 'memory']) {
resourcePerCell[t] = resourceUnits[podSpec.gpuType][t];
Expand Down Expand Up @@ -222,7 +222,7 @@ const hivedValidate = async (protocolObj, username) => {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: ${taskRole} requests ${gpu} GPU, ${cpu} CPU, ${memoryMB}MB memory; ` +
`Taskrole ${taskRole} requests ${gpu} GPU, ${cpu} CPU, ${memoryMB}MB memory; ` +
`sku allows ${resourcePerCell.gpu} GPU, ${resourcePerCell.cpu} CPU, ${resourcePerCell.memory}MB memory per cell.`
);
}
Expand All @@ -234,7 +234,7 @@ const hivedValidate = async (protocolObj, username) => {
throw createError(
'Bad Request',
'InvalidProtocolError',
`Hived error: exceed ${cellQuota} GPU quota in ${virtualCluster} VC.`
`Exceed ${cellQuota} GPU quota in ${virtualCluster} VC.`
);
}

Expand Down

0 comments on commit 279c0e5

Please sign in to comment.