What Loki costs to own, and which edition fits
Loki carries no license fee. Grafana Loki is licensed under AGPLv3, so the cost of running it is recurring infrastructure, operating time and any support you buy. Your comparison sets three paths side by side: self-hosted AGPLv3 Loki, Grafana Enterprise Logs (GEL) and hosted logging. Each path charges for different things, but in every case the bill grows with daily log volume and retention, so model those two inputs before comparing options.
Self-hosted Loki moves cost into object storage, compute for the write, read and backend components, and the platform time to run upgrades, backups and access controls. Loki has no built-in authentication, so you also pay to build and operate an authenticating reverse proxy. GEL adds token authorization, label-based access control and cross-tenant queries under a Grafana Labs agreement; confirm pricing and terms directly with Grafana Labs. Hosted logging shifts storage, compute and much of the operating work to the provider, so confirm how ingestion, retention and query volume are metered.
Four events should prompt a fresh cost review: a hosted bill that grows faster than your log volume, a planned move off Promtail, a deployment that relies on simple scalable mode, and an upgrade path toward Loki 4.0. Promtail reached end of life on March 2, 2026, so any Promtail estate needs migration to Grafana Alloy, which adds shipper effort. Simple scalable mode is deprecated and will be removed in Loki 4.0, so a deployment that depends on it needs a target architecture and an updated cost model first.
The short answer: most teams first model the self-hosted option in-house, because the largest unknowns are internal platform hours and retention decisions that only your organization can set. Bring in a partner when internal staff cannot cover the build and run work, or when you want a fixed, written scope to compare against a hosted quote. Ask any provider for line items covering infrastructure, one-time build, recurring operations and support tiers, each tied to your daily volume and retention.
What a cost of ownership review covers
A cost of ownership review turns your log volume, retention needs and staffing into a written model you can compare against Grafana Enterprise Logs and hosted logging. The scope is agreed in writing before work starts.
Volume and retention inputs
We document daily ingest by source, peak-to-average ratios, compression assumptions, and retention periods per tenant or log stream. Retention in Loki is off by default and set by the compactor, so the retention schedule must be written down before any storage figure is calculated. Confirm each input with the team that owns the source.
Cost model across infrastructure
The model covers object storage, compute for each deployment mode, and the network and backup costs around them. Each line shows whether it is paid to a cloud provider, a hardware vendor or your own staff. Ask for the unit prices and the assumptions behind them, not a single total.
People and support effort
This card estimates the internal platform time to run Loki, upgrades, on-call coverage and Grafana access administration, and where a retainer would replace some of that time. Staffing assumptions drive most of the recurring cost, so each role and its share of time should be stated explicitly.
Edition comparison
The review compares open-source Loki under AGPLv3 with Grafana Enterprise Logs, which adds token authorization, label-based access control and cross-tenant queries and is in long-term support until February 1, 2029. It also compares hosted logging. Each option is scored against the same requirements so the differences in features and cost are visible side by side.
Recurring cost schedule
We build a schedule that separates one-time work, such as initial deployment and migration, from recurring items such as storage growth, license fees, support and upgrade effort. The schedule shows when each recurring cost starts and how it changes as volume or tenant count grows. Ask for it to be reviewed against your budget cycle.
Deliverables and written scope
You receive a model workbook, an assumptions register, a decision memo and a quote basis that shows how the scope maps to the price. Scope, deliverables and timelines are agreed in writing before work starts. Confirm with us which assumptions you own and which we validate.
What drives the cost of owning Loki
Each driver that moves the total cost of a self-hosted Loki stack, why it moves that cost, and how to keep it under control.
| Driver | Why it changes cost | How to keep it down |
|---|---|---|
| Daily log volume | Ingest rate sets write load, chunk and index storage growth, and which deployment mode you need. | Drop noisy debug output at the shipper, set volume budgets per team, and review top sources each month. |
| Deployment mode | Monolithic suits up to about 20 GB per day, simple scalable approaches 1 TB per day, and microservices serve very large clusters, each adding more nodes to run. | Start with the smallest mode that fits current volume, and plan the move off simple scalable before Loki 4.0 removes it. |
| Retention period | Storage grows with how long chunks are kept, and retention is disabled by default, so logs accumulate until you enable it. | Set retention_enabled on purpose, use per-tenant or per-stream rules (at least 24 hours each), and review periods on a fixed schedule. |
| Tenants and teams | Each tenant adds auth mapping, dashboards and query limits to maintain, and multi-tenancy requires auth_enabled and X-Scope-OrgID handling. | Combine teams into fewer tenants where isolation allows, and automate tenant onboarding through your identity provider. |
| Object store choice | Object storage bills for capacity, requests and data transfer, and pricing differs between Amazon S3, Google Cloud Storage and Azure Blob Storage. | Use the store your cloud account already operates, and compare request and transfer costs in your region before sizing. Confirm S3-compatible options with Grafana's docs first. |
| Shipper estate | Every host, container and node runs a collector, so agent count drives CPU, memory and upkeep work. | Standardize on Grafana Alloy. Promtail reached end of life on March 2, 2026, so move remaining agents to Alloy. |
| Grafana Enterprise Logs features | Label-based access control and cross-tenant queries are commercial GEL features, so their license cost recurs each term. | Use open-source Loki with an authenticating proxy if access rules are simple, and request GEL pricing in writing against a listed feature set. |
| Compliance-driven retention | Regulations or internal policy can set minimum keep periods and deletion evidence, which lengthens retention and adds audit work. | Confirm required periods with your auditor or legal team, then scope long retention only to the streams that need it. |
How the cost model is built
The sequence turns measured log volumes, retention rules and lifecycle dates into a one-time and recurring cost model that finance and security have reviewed and can maintain.
Collect current log volumes and sources
List every log source, its average and peak daily volume, and the shippers that collect it, including any Promtail agents still running. Ask platform owners for measured figures from current tooling where they exist. The output is a source register with daily volume per source, which feeds every later calculation.
Fix retention and tenancy assumptions
Agree retention periods by log class and by tenant, including streams that must be kept longer or deleted sooner. Loki keeps logs indefinitely until retention is enabled, so the model must state the retention setting explicitly. The output is a signed retention and tenant table that sets the stored volume used in the storage line.
Choose deployment mode and edition options
Match the daily volume from the first step to a deployment mode, using the Loki documentation's ranges for monolithic and simple scalable modes. Simple scalable mode is removed in Loki 4.0, so price a later migration if you choose it. Add AGPLv3 Loki, Grafana Enterprise Logs and any hosted option. The output is one priced scenario per shortlisted option.
Model one-time and recurring costs
Split costs into one-time items such as design, build, migration and initial testing, and recurring items such as object storage, compute, network, backups, licenses and internal staff time. Show who pays each line: your cloud or hardware budget, your team, or a partner under a fixed-scope written quote. The output is a workbook covering the full planning horizon, with each assumption visible.
Check version and lifecycle dates
Confirm the Loki release you will pin and the support dates of every component in the design. Grafana Enterprise Logs is in long-term support from version 3.6 and reaches end of life on February 1, 2029. Promtail reached end of life on March 2, 2026, and simple scalable mode is removed in Loki 4.0. The output is a lifecycle table that flags costs falling due inside the horizon.
Review the model with finance and security
Walk finance through the cost lines and security through data location, retention, and access controls, since these drive storage, proxy and audit-related effort. Record each question and the change it caused. The output is a reviewed model with an open-items list and a named owner for every item.
Hand over the cost model
Deliver the workbook, assumptions log and lifecycle table to the budget owner, with a list of figures to refresh at each budget cycle. Attach the questions to put to any partner so quotes show which items are fixed-scope, which are included in support, and which recur. The output is a signed handover that the team can update without the original authors.
Where ownership estimates usually go wrong
Most Loki cost models miss recurring work and risks that sit outside the software itself. The points below show where estimates tend to fail and how to check each one before the model is approved.
Treating the license as the whole cost
Loki is licensed under AGPLv3, so the license itself adds no fee, but it is only one line in the model. Object storage, compute, retention, the reverse proxy, Grafana access and internal platform time all recur. Ask legal counsel to review AGPLv3 obligations for any modifications you plan to host for users, and compare the result with Grafana Enterprise Logs if you need commercial features.
Leaving retention disabled
Retention is disabled by default, so logs stay in object storage until retention_enabled is set. Model storage growth from explicit retention periods, set globally, per tenant or per stream selector, with each rule at least 24 hours. Confirm that the compactor applies deletion in a test environment before you rely on the figure in the budget.
Sizing storage from average volume
Average daily volume hides the bursts that fill storage, overload write paths and slow queries during incidents. Size storage and compute from peak volume measured over a representative period, including release weeks and incident periods. Check the result against the deployment mode guidance, which lists monolithic mode for up to about 20 GB of logs per day and simple scalable mode approaching 1 TB per day.
Ignoring the authentication layer
Loki has no built-in authentication, and its documentation requires an authenticating reverse proxy such as nginx, HAProxy, OAuth2 Proxy or Pomerium in front of it. Budget for that proxy's deployment, TLS certificates, session handling and identity provider integration. Also budget for its patching and on-call coverage, and confirm which team owns it.
Underestimating upgrade and on-call time
Each Loki upgrade needs schema and index checks, Helm chart changes, query tests and retention verification. Estimate that work for every planned release and for the year, not as a single build task. Add on-call time for ingestion gaps, compactor faults and object storage errors, and confirm who covers those alerts outside business hours.
Building on components leaving support
Promtail reached end of life on March 2, 2026, with no further updates or commercial support, so new estimates should assume Grafana Alloy from the start. Simple scalable mode is deprecated and will be removed with Loki 4.0, so a three-target design needs its own migration line item. Confirm which deployment mode will remain supported through your planning horizon.
One-time and recurring costs after you choose Loki
This table separates one-time setup costs from recurring operating costs, names who pays for each line and what drives it, and lists what to request in a quote so options can be compared on the same basis.
| Cost component | Who pays and when | What drives it | What to request in a quote |
|---|---|---|---|
| One-time: design and edition evaluation | Your team or the partner, once, before build starts | Deployment mode choice, tenancy model, number of log sources, and whether Grafana Enterprise Logs is evaluated alongside open-source Loki | A written design with the chosen deployment mode, the edition comparison, and the decisions that remain open |
| One-time: shipper migration | Your platform team or the partner, once per source group | Number of hosts and applications, existing agents, and the move from Promtail to Grafana Alloy, since Promtail reached end of life on March 2, 2026 | A list of sources, the shipper for each, and the test method used to confirm that labels and parsing match |
| One-time: initial build and dashboards | Your platform team or the partner, once | Environment count, object storage setup, reverse proxy and identity integration, and the number of dashboards and alert rules to rebuild | Named deliverables for each environment, dashboards and alert rules in scope, and the acceptance tests |
| One-time: security review | Your security team, a partner, or both, once and again after major changes | Authentication in front of Loki, tenant isolation via X-Scope-OrgID, object storage permissions, and TLS between components | The hardening checklist applied, the evidence produced, and the findings that stay open with owners |
| Recurring: object storage capacity and requests | Your cloud account or data center budget, monthly or per billing period | Daily ingest volume, retention periods set in the compactor, index size, and the number of read requests from queries | The storage class, lifecycle rules, expected request patterns, and the provider's pricing for capacity and requests |
| Recurring: compute for read, write and backend | Your infrastructure budget, continuous | Deployment mode (monolithic, simple scalable, or microservices), query load, ingest rate, and the replica count for each target | Resource requests per component, autoscaling rules, and the sizing basis tied to daily volume |
| Recurring: backups, Grafana hosting, reverse proxy and identity | Your infrastructure budget and your identity platform, continuous | Backup frequency and copy count for configuration and retention rules, the Grafana instances serving users, and the proxy and SSO licensing or hosting | Backup schedule and restore test records, Grafana sizing, and whether identity licensing is included |
| Recurring: monitoring, upgrades and internal platform time | Your platform team, continuous; upgrades recur with each Loki release | Number of Loki releases applied per year, Helm chart changes, index and schema changes, and query regression testing after upgrade | Upgrade plan per release, the testing scope, and the hours or roles assumed for on-call and monitoring |
| Recurring: vendor or partner support and Grafana Enterprise Logs | Partner retainer or support contract, and Grafana Labs for the GEL subscription if chosen | Coverage hours, response scope, and for GEL: token authorization, label-based access control and cross-tenant queries, with long-term support through February 1, 2029 | Support scope in writing, the GEL subscription terms and term length, and what is excluded from each |
How the comparison changes by starting point
With AGPLv3 Loki, there is no license fee, so the recurring cost is the infrastructure and people you run yourselves. You carry every line item.
- Object storage: you pay for compressed chunks and the index in your object store, and the bill grows with retention.
- Compute: you size the deployment mode you choose and pay for it continuously.
- Upgrades: you plan and test each version move. The latest release listed is v3.7.8, published September 17, 2026, so confirm the target version before you budget.
- Support: no vendor support contract is included. Internal staff or a partner retainer covers incidents.
- Modifications: if you change Loki and offer it to users over a network, review your AGPLv3 source-offer duties with counsel before you ship changes.
Grafana Enterprise Logs keeps the Loki engine and adds token authorization, label-based access control and cross-tenant queries. The subscription is a recurring fee on top of the infrastructure costs you already carry.
- Access features: token authorization and label-based access control may reduce the custom proxy logic you would otherwise write. Confirm the exact scope with Grafana Labs.
- Cross-tenant queries: central security or platform teams can query across tenants without maintaining separate access paths.
- Support horizon: long-term support covers version 3.6 through February 1, 2029, which sets a planning window for upgrades.
- Infrastructure: object storage, compute and operations stay on your side of the bill.
- Subscription terms: ask Grafana Labs for scope, term, support levels and renewal conditions in writing before you compare totals.
A hosted logging service moves storage, compute and operations into a provider fee. The comparison then turns on data terms and exit conditions, which you should confirm before signing.
- Data residency: confirm the regions where logs are stored and processed, and whether backups or replicas leave that region.
- Retention pricing: check how the provider prices longer retention, since extended periods can multiply the bill.
- Egress and exit: ask what it costs to export logs and how long a full export takes.
- Exit effort: query languages, dashboards and alert rules differ between platforms, so budget migration work at exit.
- Comparison basis: price self-hosting and hosted options over the same retention period, including your staff time.
In isolated or sovereign networks, object storage runs on premises and updates arrive by controlled transfer. Those two changes raise the recurring cost compared with a connected environment.
- Object storage: you need an S3-compatible store you operate. Loki's documentation says MinIO and the local filesystem are supported but not typically recommended for production, so confirm your choice against the current storage guidance.
- Update delivery: each release, patch and chart update needs a staged transfer, scanning and review, which adds staff time per cycle.
- Mirrored artifacts: container images and Helm charts must be mirrored inside the network before deployment.
- Capacity: retention drives disk growth, so plan capacity per tenant and keep headroom for compaction.
- Quotes: ask providers to price mirroring, offline patching and retention capacity as separate lines.
What you need to provide for a cost review
A Loki cost review is only as accurate as the inputs behind it, so gather these before anyone builds a model. Provide daily log volume by source, with peak figures alongside averages, because the busiest hours set the sizing. Add retention requirements for each log class and the regulator or internal policy behind it. List the number of teams and tenants, plus your current hosted or Elasticsearch-based bills, including contract terms, renewal dates and minimum commitments.
Settle three decisions before pricing, because each one changes the cost structure. First, choose the deployment mode. Loki documents monolithic mode for up to about 20 GB of logs per day. Simple scalable mode is being deprecated and will be removed with Loki 4.0, so a plan built on it carries a later migration. Second, decide whether to adopt Grafana Enterprise Logs, which adds token authorization, label-based access control and cross-tenant queries, with long-term support through February 1, 2029. Third, confirm who will run the authenticating reverse proxy and tenant mapping that Loki requires.
Also document the platform you already run. Name the object store, such as Amazon S3, Google Cloud Storage, Azure Blob Storage or an S3-compatible system, along with its contracts and current storage, request and egress charges. Describe the Kubernetes or VM platform that will host Loki and its unit costs for compute. Record your Grafana instance, identity provider and log shippers. Promtail reached end of life on March 2, 2026, so flag any agents still using it as migration work in the estimate.
Name one owner each for finance, security and platform operations, and agree who approves the cost model. Finance should confirm which costs sit in cloud accounts, which in existing licenses and which in internal headcount. Security should own retention and deletion rules for sensitive logs. Platform operations should estimate the internal time to run the stack. Ask any implementation partner to quote against this written input list and to state the assumptions used for any missing input.
How DeployOpen builds an ownership model
DeployOpen helps you compare the cost of running self-hosted Loki with Grafana Enterprise Logs and hosted logging, then turns that comparison into a written model your finance and platform teams can review. The model separates one-time work from recurring cost, in infrastructure you control.
How we scope
- Confirm daily log volume, retention periods per tenant or stream, and the deployment mode the volume points toward
- List object storage, compute, shipper and Grafana access assumptions, each with an owner and a source for its price
- Record the Enterprise Logs and hosted logging alternatives you want compared, including the features each would need
How we build
- Build a cost workbook with one-time items such as design, migration and hardening separated from recurring items
- Model storage growth under your retention rules, including the effect of retention being disabled by default
- Test assumptions against a sample of your log volume and document where the estimate is least certain
How we hand over
- Deliver the cost model as an editable workbook with a written summary of assumptions and open questions
- Provide a checklist of questions to put to each vendor or hosting provider when requesting a quote
- Walk your finance, platform and security owners through the model so they can change inputs without us
How we support
- Offer a review of the model when licensing, Loki releases or your log volume change, under a retainer if you want one
- Track the Promtail end-of-life and Loki 4.0 simple scalable removal as items that may change recurring costs
- Keep the model aligned with each fixed-scope deliverable so changes to scope are agreed in writing first
