Execution Modes

Execution modes determine how deployments run across multiple targets.

Available Modes

Mode	Description	Best For
Parallel	All targets simultaneously	Independent services
Sequential	One target at a time	Stateful services
Rolling	Batches of targets	Large deployments

Parallel Execution

All targets receive commands at the same time:

Characteristics

Property	Value
Speed	Fastest
Resource usage	High (all targets active)
Risk	Higher (all targets affected at once)
Rollback complexity	Higher

When to Use

Stateless services
Independent targets
Speed is critical
Targets don’t share state

When to Avoid

Database servers
Services with shared state
When gradual rollout is required
Limited network bandwidth

Sequential Execution

Targets are updated one at a time:

Characteristics

Property	Value
Speed	Slowest
Resource usage	Low (one target active)
Risk	Lower (issues caught early)
Rollback complexity	Lower

When to Use

Database servers
Primary/replica pairs
Ordered dependencies
Careful rollouts

When to Avoid

Large number of targets (takes too long)
Time-sensitive deployments
Independent services

Rolling Execution

Targets are updated in batches:

Configuration

Setting	Description	Required?
Batch Size	Targets per batch	Yes — minimum 1; there is no default. Lens prefills 2.

Characteristics

Property	Value
Speed	Medium
Resource usage	Medium
Risk	Balanced
Rollback complexity	Medium

Batch Size Strategy

Targets	Recommended Batch Size
4-10	2
10-50	5-10
50-100	10-20
100+	10-25 (5-10% of total)

When to Use

Large deployments
Maintaining service availability
Gradual rollouts
Canary deployments

Configuring Execution Mode

In the UI

┌─────────────────────────────────────────────────────────┐
│  Execution Configuration                                │
├─────────────────────────────────────────────────────────┤
│                                                         │
│  Mode:                                                  │
│  ○ Parallel      - All targets at once                  │
│  ○ Sequential    - One target at a time                 │
│  ● Rolling       - Targets in batches                   │
│                                                         │
│  Batch Size:     [3                              ]      │
│                                                         │
└─────────────────────────────────────────────────────────┘

Via CLI

# Parallel
mantisctl deployment create --solution-id <id> --mode parallel

# Sequential
mantisctl deployment create --solution-id <id> --mode sequential

# Rolling, two targets at a time
mantisctl deployment create --solution-id <id> --mode rolling:2

--mode takes the batch size inline as rolling:N (N of at least 1); the same deployment is expressed through the API as the target_mode field.

Target Order

A deployment runs the targets its request names, in the order named. Sequential and rolling modes follow that order; parallel mode starts all targets together and reports results as they complete.

Promotions and rollbacks resolve their targets from the environment instead, ordered by each assignment’s priority (-100 to 100, default 50, higher first):

Target	Priority	Execution Order
web-prod-01	80	First
web-prod-02	70	Second
web-prod-03	60	Third
db-prod-01	10	Last

See Target Selection.

Failure Handling by Mode

Parallel + Stop on Failure

All targets start
On failure, no new steps start
Running steps complete
Deployment marked failed

Sequential + Stop on Failure

First failure stops entire deployment
Remaining targets skipped
Easy to identify failing target

Rolling + Stop on Failure

Current batch completes
Remaining batches skipped
Partial deployment state

Comparison Table

Aspect	Parallel	Sequential	Rolling
Deployment time	Fastest	Slowest	Medium
Risk exposure	All at once	Gradual	Balanced
Availability during deploy	Potentially zero	Gradual reduction	Controlled
Rollback scope	All targets	Single target	Batch
Monitoring complexity	High	Low	Medium
Best for	Stateless	Stateful	Large scale

Best Practices

1. Match Mode to Service Type

Service Type	Recommended Mode
Stateless API	Parallel or Rolling
Database	Sequential
Cache layer	Rolling
Load balancer	Sequential
Worker queue	Rolling

2. Use Rolling for Production

Even for stateless services:

Rolling (batch size: 25%)
├── Catch issues early
├── Maintain availability
└── Allow monitoring between batches

3. Consider Deployment Windows

Mode	Suitable Window
Parallel	Short windows
Sequential	Long windows
Rolling	Medium windows

4. Plan for Rollback

Think about how you’ll rollback:

Mode	Rollback Strategy
Parallel	Rollback all
Sequential	Rollback completed
Rolling	Rollback by batch

Monitoring by Mode

Parallel

Watch for:

Overall completion percentage
Any failures across all targets
Resource utilization spikes

Sequential

Watch for:

Current target progress
Cumulative success rate
Time per target

Rolling

Watch for:

Current batch progress
Batch completion rate
Issues between batches

Troubleshooting

Rolling Deployment Too Slow

Cause: Batch size too small

Solution: Increase the batch size

Sequential Deployment Stuck

Cause: Single target blocking

Solution: Check target status, consider skipping or fixing

Parallel Overwhelming Resources

Cause: Too many concurrent operations

Solution: Switch to rolling mode with appropriate batch size

Next Steps

Creating Deployments - Create new deployments
Target Selection - Choose targets
Monitoring - Track progress
Overview - Return to deployments overview