add "Understanding the Metrics" section explaining three types: - per-worker rates (multiply by active workers for total throughput) - per-region averages (do NOT multiply by worker count) - system totals (cumulative across all workers) enhance metric descriptions with: - explicit calculation examples showing which worker counts to use - warnings about when NOT to multiply by worker count - cross-references to relevant sections add "Capacity Planning" section with: - step-by-step throughput capacity calculations - diagnostic guidance for common scenarios - worked examples for estimating required vs actual capacity addresses confusion from PR #11034 comments about when to multiply metrics by worker count and how to interpret per-worker rates
9.5 KiB
Provide Stats
The ipfs provide stat command gives you statistics about your local provide
system. This file provides a detailed explanation of the metrics reported by
this command.
Understanding the Metrics
The statistics are organized into three types of measurements:
Per-worker rates
Metrics like "CIDs reprovided/min/worker" measure the throughput of a single worker processing one region. To estimate total system throughput, multiply by the number of active workers of that type (see Workers stats).
Example: If "CIDs reprovided/min/worker" shows 100 and you have 10 active periodic workers, your total reprovide throughput is approximately 1,000 CIDs/min.
Per-region averages
Metrics like "Avg CIDs/reprovide" measure properties of the work units (keyspace regions). These represent the average size or characteristics of a region, not a rate. Do NOT multiply these by worker count.
Example: "Avg CIDs/reprovide: 250,000" means each region contains an average of 250,000 CIDs that get reprovided together as a batch.
System totals
Metrics like "Total CIDs provided" are cumulative counts since node startup. These aggregate all work across all workers over time.
Connectivity
Status
Current connectivity status (online, disconnected, or offline) and when
it last changed (see provide connectivity
status).
Queues
Provide queue
Number of CIDs waiting for initial provide, and the number of keyspace regions they're grouped into.
Reprovide queue
Number of regions with overdue reprovides. These regions missed their scheduled
reprovide time and will be processed as soon as possible. If decreasing, the
node is recovering from downtime. If increasing, either the node is offline or
the provide system needs more workers (see
Provide.DHT.MaxWorkers
and
Provide.DHT.DedicatedPeriodicWorkers).
Schedule
CIDs scheduled
Total CIDs scheduled for reprovide.
Regions scheduled
Number of keyspace regions scheduled for reprovide. Each CID is mapped to a specific region, and all CIDs within the same region are reprovided together as a batch for efficient processing.
Avg prefix length
Average length of binary prefixes identifying the scheduled regions. Each keyspace region is identified by a binary prefix, and this shows the average prefix length across all regions in the schedule. Longer prefixes indicate the keyspace is divided into more regions (because there are more DHT servers in the swarm to distribute records across).
Next region prefix
Keyspace prefix of the next region to be reprovided.
Next region reprovide
When the next region is scheduled to be reprovided.
Timings
Uptime
How long the provide system has been running since Kubo started, along with the start timestamp.
Current time offset
Elapsed time in the current reprovide cycle, showing cycle progress (e.g., '11h' means 11 hours into a 22-hour cycle, roughly halfway through).
Cycle started
When the current reprovide cycle began.
Reprovide interval
How often each CID is reprovided (the complete cycle duration).
Network
Avg record holders
Average number of provider records successfully sent for each CID to distinct DHT servers. In practice, this is often lower than the replication factor due to unreachable peers or timeouts. Matching the replication factor would indicate all DHT servers are reachable.
Note: this counts successful sends; some DHT servers may have gone offline afterward, so actual availability may be lower.
Peers swept
Number of DHT servers to which we tried to send provider records in the last reprovide cycle (sweep). Excludes peers contacted during initial provides or DHT lookups.
Full keyspace coverage
Whether provider records were sent to all DHT servers in the swarm during the last reprovide cycle. If true, peers swept approximates the total DHT swarm size over the last reprovide interval.
Reachable peers
Number and percentage of peers to which we successfully sent all provider records assigned to them during the last reprovide cycle.
Avg region size
Average number of DHT servers per keyspace region.
Replication factor
Target number of DHT servers to receive each provider record.
Operations
Ongoing provides
Number of CIDs and regions currently being provided for the first time. More CIDs than regions indicates efficient batching. Each region provide uses a burst worker.
Ongoing reprovides
Number of CIDs and regions currently being reprovided. Each region reprovide uses a periodic worker.
Total CIDs provided
Total number of provide operations since node startup (includes both provides and reprovides).
Total records provided
Total provider records successfully sent to DHT servers since startup (includes reprovides).
Total provide errors
Number of failed region provide/reprovide operations since startup. Failed regions are automatically retried unless the node is offline.
CIDs provided/min/worker
Average rate of initial provides per minute per worker during the last reprovide cycle (excludes reprovides). Each worker handles one keyspace region at a time, providing all CIDs in that region. This measures the throughput of a single worker only.
To estimate total system provide throughput, multiply by the number of active burst workers shown in Workers stats (Burst > Active).
Note: This rate only counts active time when initial provides are being processed. If workers are idle, actual throughput may be lower.
CIDs reprovided/min/worker
Average rate of reprovides per minute per worker during the last reprovide cycle (excludes initial provides). Each worker handles one keyspace region at a time, reproviding all CIDs in that region. This measures the throughput of a single worker only.
To estimate total system reprovide throughput, multiply by the number of active periodic workers shown in Workers stats (Periodic > Active).
Example: If this shows 100 CIDs/min and you have 10 active periodic workers, your total reprovide throughput is approximately 1,000 CIDs/min.
Note: This rate only counts active time when regions are being reprovided. If workers are idle due to network issues or queue exhaustion, actual throughput may be lower.
Region reprovide duration
Average time to reprovide all CIDs in a region during the last cycle.
Avg CIDs/reprovide
Average number of CIDs per region during the last reprovide cycle.
This measures the average size of a region (how many CIDs are batched together), not a throughput rate. Do NOT multiply this by worker count.
Combined with Region reprovide duration, this helps estimate per-worker throughput: dividing Avg CIDs/reprovide by Region reprovide duration gives CIDs/min/worker.
Regions reprovided (last cycle)
Number of regions reprovided in the last cycle.
Workers
Active workers
Number of workers currently processing provide or reprovide operations.
Free workers
Number of idle workers not reserved for periodic or burst tasks.
Workers stats
Breakdown of worker status by type (periodic for scheduled reprovides, burst for initial provides). For each type:
- Active: Currently processing operations (use this count when calculating total throughput from per-worker rates)
- Dedicated: Reserved for this type
- Available: Idle dedicated workers + free workers
- Queued: 0 or 1 (workers acquired only when needed)
The number of active workers determines your total system throughput. For example, if you have 10 active periodic workers, multiply CIDs reprovided/min/worker by 10 to estimate total reprovide throughput.
See provide queue and reprovide queue for regions waiting to be processed.
Max connections/worker
Maximum concurrent DHT server connections per worker when sending provider records for a region.
Capacity Planning
Estimating if your system can keep up with the reprovide schedule
To check if your provide system has sufficient capacity:
-
Calculate required throughput:
- Required CIDs/min = CIDs scheduled / (Reprovide interval in minutes)
- Example: 67M CIDs / (22 hours × 60 min) = 50,758 CIDs/min needed
-
Calculate actual throughput:
- Actual CIDs/min = CIDs reprovided/min/worker × Active periodic workers
- Example: 100 CIDs/min/worker × 256 active workers = 25,600 CIDs/min
-
Compare:
- If actual < required: System is underprovisioned, increase MaxWorkers or DedicatedPeriodicWorkers
- If actual > required: System has excess capacity
- If Reprovide queue is growing: System is falling behind
Understanding worker utilization
- High active workers with growing reprovide queue: Need more workers or network connectivity is limiting throughput
- Low active workers with non-empty reprovide queue: Workers may be waiting for network or DHT operations
- Check Reachable peers to diagnose network connectivity issues