Skip to content

config-envvars.md

Latest commit

 

History

History
327 lines (257 loc) · 33.7 KB

config-envvars.md

File metadata and controls

327 lines (257 loc) · 33.7 KB
Raw

Environment Variables

Canonical reference for every OPENSEARCH_GO_* runtime environment variable. Each variable is read once at client initialization and is immutable after. Environment variable values override their programmatic configuration equivalents.

Why so many environment variables?

The client exposes a large number of operational dials — routing behavior, discovery cadence, overload thresholds, connection-pool sizing, partial-failure handling — and every one of them is configurable in code (opensearch.Config, opensearchapi.Config, and RouterOptions). Each environment variable is an override for one of those programmatic settings, so an operator can retune a deployment to meet an organization's operational needs without changing source or rebuilding the binary. The variables are not an alternative configuration system; they are deploy-time escape hatches over the in-code defaults.

Quick reference

Every runtime variable, its default, and a one-line summary. Use this as the table of contents — each variable links to its detailed entry below.

Variable Default Summary
OPENSEARCH_URL unset Seed addresses
OPENSEARCH_GO_REQUEST_TIMEOUT 0 (none) Per-attempt timeout
OPENSEARCH_GO_DNS_CACHE_REFRESH 60s Client-side DNS cache refresh
OPENSEARCH_GO_DNS_DIAL_TIMEOUT 30s DNS-cache dialer dial timeout
OPENSEARCH_GO_DNS_KEEP_ALIVE 30s DNS-cache dialer keep-alive
OPENSEARCH_GO_DNS_TIMEOUT 10s DNS-cache per-lookup refresh timeout
OPENSEARCH_GO_ROUTER true Auto-construct DefaultRouter
OPENSEARCH_GO_ROUTING_CONFIG all enabled Shard-exact and adaptive MCSR toggles
OPENSEARCH_GO_SHARD_COST defaults Shard cost multipliers
OPENSEARCH_GO_SHARD_REQUESTS true (5:256) Adaptive MCSR bounds
OPENSEARCH_GO_DISCOVERY_CONFIG all enabled Discovery call toggles
OPENSEARCH_GO_FALLBACK true Seed URL fallback
OPENSEARCH_GO_VERIFY_DEAD_AFTER 15m Zombie-resurrection expiry window
OPENSEARCH_GO_NODE_STATS_INTERVAL auto Stats polling interval
OPENSEARCH_GO_OVERLOADED_HEAP_THRESHOLD 85 JVM heap overload threshold
OPENSEARCH_GO_OVERLOADED_BREAKER_RATIO 0.90 Breaker ratio overload threshold
OPENSEARCH_GO_ACTIVE_LIST_CAP auto Max active connections per pool
OPENSEARCH_GO_STANDBY_ROTATION_INTERVAL 0 (use discovery interval) Standby rotation interval
OPENSEARCH_GO_STANDBY_ROTATION_COUNT 1 Standby rotations per cycle
OPENSEARCH_GO_STANDBY_PROMOTION_CHECKS 3 Health checks before promotion
OPENSEARCH_GO_DEFAULT_CLIENT_TTL 16m Default-client cache idle eviction
OPENSEARCH_GO_DEBUG false Debug logging
OPENSEARCH_GO_ERROR_MASK report all (v5+) Partial-failure category mask
OPENSEARCH_GO_POLICY_* all enabled Per-policy disable (10 variables)
OPENSEARCH_GO_POLICY_DUMP false Dump router policy tree (debug-gated)

Build, test, and code-generation variables (not read by the client at runtime) are listed under Build, test, and development.

Parsing rules

Boolean variables. Parsed with Go's strconv.ParseBool. Truthy values: 1, t, T, TRUE, true, True. Falsy values: 0, f, F, FALSE, false, False. Empty, unset, or unparseable values fall back to the documented default.

Duration variables. Accept a Go time.ParseDuration string (e.g. 500ms, 2s, 1m30s) or a bare integer/float number of seconds.

The "See also" column links to the guide section that explains the concept the variable tunes. This document owns the value table; the linked section owns the explanation. No concept is described in two places.

Connection

Variable Accepted values Default Meaning See also
OPENSEARCH_URL Comma-separated URL list (e.g. https://a:9200,https://b:9200) unset Seed addresses used by NewClient when no Addresses are set programmatically. opensearchapi Client Creation; Security: Credential Management
OPENSEARCH_GO_DNS_CACHE_REFRESH Duration or seconds 60s Client-side DNS cache refresh interval. Resolved addresses are re-resolved on this cadence; if the resolver becomes briefly unreachable, the last-known-good address is served until it recovers, so a transient DNS outage does not fail requests to already-resolved hosts. 0 or unset = default (60s); negative = disable caching; positive = explicit interval. Installed only on the built-in transport; a caller-supplied Transport is never modified. Because Go's resolver does not expose record TTLs, this is a re-resolution cadence, not a per-record TTL. Overrides Config.DNSCacheRefresh. opensearchapi Client Creation
OPENSEARCH_GO_DNS_DIAL_TIMEOUT Duration or seconds 30s Dial timeout of the net.Dialer behind the DNS cache. 0 or unset = default (30s); negative = no dial timeout; positive = explicit timeout. Only applies when the cache is installed (no custom Transport). Overrides Config.DNSDialTimeout. opensearchapi Client Creation
OPENSEARCH_GO_DNS_KEEP_ALIVE Duration or seconds 30s Keep-alive interval of the net.Dialer behind the DNS cache. 0 or unset = default (30s); negative = disable keep-alive probes; positive = explicit interval. Only applies when the cache is installed (no custom Transport). Overrides Config.DNSKeepAlive. opensearchapi Client Creation
OPENSEARCH_GO_DNS_TIMEOUT Duration or seconds 10s Per-lookup timeout applied to each DNS cache refresh resolution. Refresh lookups run sequentially on a single goroutine, so this bounds how long one stuck resolution can stall a refresh tick. 0 or unset = default (10s); negative = no per-lookup timeout; positive = explicit timeout. Only applies when the cache is installed (no custom Transport). Overrides Config.DNSTimeout. opensearchapi Client Creation

Routing

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_ROUTER Bool true (router auto-constructed) Auto-construct NewDefaultRouter() when no programmatic Config.Router is set. Set false or 0 to suppress. A programmatic Config.Router always takes precedence. Routing: Quick Start; Default router injection
OPENSEARCH_GO_ROUTING_CONFIG Comma-separated flags/key=value all enabled Toggle shard-exact routing (-shard_exact) and adaptive MCSR (-adaptive_mcsr). Routing: Configuration Reference
OPENSEARCH_GO_SHARD_COST Key=value, comma-separated (or bare numeric for r:base) compile-time defaults Override shard cost multipliers used in connection scoring. Format: key=value pairs (r:base=0.95, unknown=32.0) or a bare number (sets r:base). Routing: Shard Cost Configuration
OPENSEARCH_GO_SHARD_REQUESTS Bool or min:max true (5:256) Adaptive max_concurrent_shard_requests bounds. true/false enable/disable with defaults; 10:512, 10:, :512, or bare 10 set custom bounds. Routing: Configuration Reference

Default router injection

opensearchapi.NewClient (and NewDefaultClient) inject opensearchtransport.NewDefaultRouter when the caller leaves Config.Client.Router nil, so requests are routed by node role by default. OPENSEARCH_GO_ROUTER is the opt-out:

OPENSEARCH_GO_ROUTER Behavior
unset default Router injected, auto-discovery on
true / 1 default Router injected, auto-discovery on
false / 0 injection skipped (Router stays nil), no auto-discovery
unparseable default Router injected, auto-discovery on 
// Default: NewDefaultRouter is injected.
client, _ := opensearchapi.NewClient(opensearchapi.Config{
    Client: opensearch.Config{Addresses: addrs}, // Router == nil
})

// Caller-provided Router is preserved.
custom := opensearchtransport.NewMuxRouter()
client, _ = opensearchapi.NewClient(opensearchapi.Config{
    Client: opensearch.Config{Addresses: addrs, Router: custom},
})

A caller-supplied DiscoverNodesOnStart value always wins over the env-var-driven side-effect: setting DiscoverNodesOnStart: &false keeps auto-discovery off even when OPENSEARCH_GO_ROUTER=true. For routing semantics (role awareness, AIMD, shard-cost weighting) see Request Routing; for node discovery see Node Discovery and Roles.

Discovery

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_DISCOVERY_CONFIG Comma-separated flags all enabled Skip specific discovery calls. Flags: -cat_shards, -routing_num_shards, -cluster_health, -node_stats. Routing: Configuration Reference
OPENSEARCH_GO_FALLBACK Bool true Seed URL fallback when all router pools are exhausted. Set false to disable. Routing: Configuration Reference
OPENSEARCH_GO_VERIFY_DEAD_AFTER Bool or duration/seconds 15m How long a connection proven reachable stays a blind zombie-resurrection candidate while dead. Past this window the discovery loop stops treating the node as a last-resort target until it health-checks clean again. Bool true selects the default, false disables it (a proven connection stays a candidate indefinitely); any other value is a duration (10m) or bare seconds. Seed connections are exempt. Overrides Config.VerifyDeadAfter (0 = default, <0 = disabled, >0 = explicit). Routing: Connection Pool Lifecycle

Load shedding and stats polling

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_REQUEST_TIMEOUT Duration or seconds 0 (none) Per-attempt HTTP round-trip timeout. Security: Connection Timeouts
OPENSEARCH_GO_NODE_STATS_INTERVAL Duration or seconds auto (5s-30s) Stats polling interval. 0 or unset = auto; negative = disabled. Routing: Load Shedding and Stats Polling
OPENSEARCH_GO_OVERLOADED_HEAP_THRESHOLD Integer (0-100) 85 JVM heap percentage at or above which a node is marked overloaded (comparison is >=). Routing: Load Shedding and Stats Polling
OPENSEARCH_GO_OVERLOADED_BREAKER_RATIO Float (0.0-1.0] 0.90 Circuit-breaker estimated_size / limit_size ratio at or above which a node is marked overloaded (comparison is >=). Values outside (0.0, 1.0] are ignored. Routing: Load Shedding and Stats Polling

Connection pool tuning

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_ACTIVE_LIST_CAP Integer auto Max active connections per pool. 0 or unset = auto-scale with cluster size. Routing: Connection Pool Lifecycle
OPENSEARCH_GO_STANDBY_ROTATION_INTERVAL Duration or seconds 0 (use DiscoverNodesInterval) Interval between standby rotation cycles. 0 or unset inherits the node-discovery interval; negative disables rotation. Routing: Connection Pool Lifecycle
OPENSEARCH_GO_STANDBY_ROTATION_COUNT Integer 1 Standby connections rotated per cycle. Routing: Connection Pool Lifecycle
OPENSEARCH_GO_STANDBY_PROMOTION_CHECKS Integer 3 Consecutive successful health checks required to promote a standby connection to active. Routing: Connection Pool Lifecycle

Default client cache

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_DEFAULT_CLIENT_TTL Duration or seconds 16m Idle eviction window for the process-wide cache of implicitly-created default clients (opensearch.NewDefaultClient, opensearchapi.NewDefaultClient, and the client opensearchutil.NewBulkIndexer builds when none is supplied). Accepts a time.ParseDuration string (16m) or a bare number of seconds (30, 1.5). Identical default clients share one cached transport, keyed by config hash, until every holder is closed and the entry sits idle for a full window. 0 = never evict (entries live until process exit); a negative value (-1, -1s) disables caching, so every call builds a fresh client; positive = explicit window. The 16m default is the 15m AWS Lambda max timeout plus a 1m buffer, so a default client is not evicted mid-invocation across the longest possible Lambda run. Has no opensearch.Config equivalent — user-built opensearch.NewClient/opensearchapi.NewClient clients never enter the cache. opensearchapi Client Creation

Debug and diagnostics

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_DEBUG Bool false Enable verbose internal logging to stderr (routing decisions, discovery, pool operations). Equivalent to setting EnableDebugLogger: true in opensearch.Config. USER_GUIDE.md Debugging

Error masking

Variable Accepted values Default Meaning See also
OPENSEARCH_GO_ERROR_MASK Comma-separated +/- tokens version-dependent (v5+: report all) Mask (ignore) specific categories of partial-failure errors. Tokens applied left-to-right starting from Config.Errors. Unrecognized tokens are silently dropped and logged when OPENSEARCH_GO_DEBUG=true. Error Handling; Error Handling

OPENSEARCH_GO_ERROR_MASK tokens

Each token is the lowercase snake_case form of the corresponding wrapper-schema name. A bare token (no prefix) is treated as + (set/mask the bit). A - prefix clears/unmasks the bit.

Token Category masked Returned by
bulk_items BulkItems Bulk, BulkStream
search_shards SearchShards Search, MSearch, MSearchTemplate, SearchTemplate, Scroll.Get, Count, CreatePIT
write_shards WriteShards Index, Document.Create, Document.Delete, Update
broadcast_shards BroadcastShards Indices.Refresh, Indices.Flush, Indices.ForceMerge, and similar broadcast endpoints
node_failures NodeFailures Cluster.Stats, Nodes.Info, Nodes.Stats, Nodes.Usage
bulk_by_scroll_failures BulkByScrollFailures Reindex, UpdateByQuery, DeleteByQuery
task_failures TaskFailures Tasks.List, Tasks.Cancel
multi_search_items MultiSearchItems MSearch, MSearchTemplate
multi_doc_items MultiDocItems MGet, MTermvectors
snapshot_create_shard_failures SnapshotCreateShardFailures Snapshot.Create (when wait_for_completion=true)
snapshot_get_shard_failures SnapshotGetShardFailures Snapshot.Get
simulate_doc_failures SimulateDocFailures Ingest.Simulate
rank_eval_failures RankEvalFailures RankEval
ingestion_shard_failures IngestionShardFailures Ingestion.Pause, Ingestion.Resume
pit_node_failures PitNodeFailures GetAllPits

Special tokens:

Token Meaning
all Set every category bit (mask everything)
empty, none, unknown Clear every category bit (mask nothing — report everything). All three are aliases for the same value.

Examples:

# Mask everything except bulk-item errors
export OPENSEARCH_GO_ERROR_MASK="+all,-bulk_items"

# Only mask search-shard failures; report every other category
export OPENSEARCH_GO_ERROR_MASK="search_shards"

# Mask everything (mimics the v4 default)
export OPENSEARCH_GO_ERROR_MASK="all"

# Report everything (the v5+ default)
export OPENSEARCH_GO_ERROR_MASK="none"

Policy overrides

Ten variables let operators disable specific routing policies at startup without code changes. All policies are enabled by default. Set a variable to false or 0 to disable all instances of that policy type. Each policy type links to its godoc.

Variable Policy type Meaning
OPENSEARCH_GO_POLICY_CHAIN PolicyChain Controls whether the chain iterates its children
OPENSEARCH_GO_POLICY_MUX MuxPolicy Controls the multiplexer that fans-out routes by pool name
OPENSEARCH_GO_POLICY_IFENABLED IfEnabledPolicy Controls conditional policy branching
OPENSEARCH_GO_POLICY_ROUTER poolRouter (unexported) Controls connection-scoring routing
OPENSEARCH_GO_POLICY_ROLE RolePolicy Controls role-filtered node selection
OPENSEARCH_GO_POLICY_ROUNDROBIN RoundRobinPolicy Controls round-robin fallback selection
OPENSEARCH_GO_POLICY_COORDINATOR CoordinatorPolicy Controls coordinating-only node routing
OPENSEARCH_GO_POLICY_NULL NullPolicy Controls the terminal no-op policy
OPENSEARCH_GO_POLICY_INDEX_ROUTER IndexRouter Controls per-index fan-out routing
OPENSEARCH_GO_POLICY_DOCUMENT_ROUTER DocRouter Controls document-ID-based shard targeting

OPENSEARCH_GO_POLICY_ROUTER targets the unexported poolRouter type, which has no godoc page; it is the connection-scoring wrapper around each role policy.

Value format

Bool (disable all instances):

OPENSEARCH_GO_POLICY_ROLE=false      # fall through to round-robin
OPENSEARCH_GO_POLICY_ROUTER=false    # plain role-based routing, no scoring

Setting true is a no-op (same as unset — all policies are enabled by default).

Overrides are applied once at startup and cannot be changed at runtime. See Routing: Policy Override Variables for the full reference.

Targeting specific policy instances (path matchers)

Instead of a bare bool, an OPENSEARCH_GO_POLICY_* value may be a comma-separated list of path=bool items that target individual nodes in the router's policy tree. The path is matched first as a regular expression, then (if that does not compile or match) as a string prefix:

# Disable one exact node: the search-pool router's search-role branch
# (see the labeled tree below to find this path)
OPENSEARCH_GO_POLICY_ROUTER=ifenabled[0].chain[0].mux[0].router[8]=false

# Disable every role node under any router via regex
OPENSEARCH_GO_POLICY_ROLE=.*router.*role.*=false

A value with no = is treated as a regex pattern that disables matching nodes (the =false is implied).

Finding the paths: the router "DOM"

Path matchers target the dot-delimited node paths the client assigns when it walks the policy tree. There is no public API that returns this tree, so to discover the paths for your router, set OPENSEARCH_GO_POLICY_DUMP=true together with OPENSEARCH_GO_DEBUG=true and read the dump from stderr at client initialization:

OPENSEARCH_GO_DEBUG=true OPENSEARCH_GO_POLICY_DUMP=true ./your-app

OPENSEARCH_GO_POLICY_DUMP writes through the debug logger, so it produces output only when OPENSEARCH_GO_DEBUG is also truthy. It does not change routing behavior; it only prints the tree.

For reference, the default router (the tree built by NewDefaultRouter when OPENSEARCH_GO_ROUTER is on) produces this 58-node tree. Each router node is labeled with the thread pool it scores for, and each role node with the role it selects, because the bare path (router[0] vs router[8]) does not say which pool or role a node serves. Your tree may differ if you supply a custom router or RouterOptions:

Router policy tree (58 nodes); target these paths with OPENSEARCH_GO_POLICY_*:
  ifenabled[0]
  ifenabled[0].chain[0]
  ifenabled[0].chain[0].mux[0]
  ifenabled[0].chain[0].mux[0].router[0]  pool=flush
  ifenabled[0].chain[0].mux[0].router[0].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[0].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[0].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[1]  pool=force_merge
  ifenabled[0].chain[0].mux[0].router[1].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[1].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[1].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[2]  pool=get
  ifenabled[0].chain[0].mux[0].router[2].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[2].ifenabled[0].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[2].ifenabled[0].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[2].ifenabled[0].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[2].ifenabled[0].role[0]  role=search
  ifenabled[0].chain[0].mux[0].router[3]  pool=management
  ifenabled[0].chain[0].mux[0].router[3].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[3].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[3].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[4]  pool=management
  ifenabled[0].chain[0].mux[0].router[4].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[4].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[4].ifenabled[0].role[0]  role=ingest
  ifenabled[0].chain[0].mux[0].router[5]  pool=management
  ifenabled[0].chain[0].mux[0].router[5].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[5].ifenabled[0].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[5].ifenabled[0].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[5].ifenabled[0].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[5].ifenabled[0].role[0]  role=search
  ifenabled[0].chain[0].mux[0].router[6]  pool=management
  ifenabled[0].chain[0].mux[0].router[6].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[6].ifenabled[0].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[6].ifenabled[0].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[6].ifenabled[0].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[6].ifenabled[0].role[0]  role=warm
  ifenabled[0].chain[0].mux[0].router[7]  pool=refresh
  ifenabled[0].chain[0].mux[0].router[7].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[7].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[7].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[8]  pool=search
  ifenabled[0].chain[0].mux[0].router[8].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[8].ifenabled[0].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[8].ifenabled[0].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[8].ifenabled[0].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[8].ifenabled[0].role[0]  role=search
  ifenabled[0].chain[0].mux[0].router[9]  pool=write
  ifenabled[0].chain[0].mux[0].router[9].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[9].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[9].ifenabled[0].role[0]  role=data
  ifenabled[0].chain[0].mux[0].router[10]  pool=write
  ifenabled[0].chain[0].mux[0].router[10].ifenabled[0]
  ifenabled[0].chain[0].mux[0].router[10].ifenabled[0].null[0]
  ifenabled[0].chain[0].mux[0].router[10].ifenabled[0].role[0]  role=ingest
  ifenabled[0].chain[0].roundrobin[0]
  ifenabled[0].router[0]
  ifenabled[0].router[0].role[0]  role=coordinating_only

Reading the tree: the mux[0] fans requests out to one router per (operation, pool) pairing; each router scores connections for its pool and delegates to an ifenabled/role subtree that selects nodes by role (with null as the give-up branch). Several routers wrap the same underlying role policy (e.g. every role=data leaf), so the same logical policy appears at multiple paths; a path matcher targets the position in the tree, not the shared instance. The labels (pool=, role=) are not part of the path — match on the path text to the left of the two-space gap.

Node lines are emitted in tree-traversal order (depth-first, siblings ordered as the matcher orders them), so the printed order is the order the override matcher walks. Override actions are also logged to stderr when OPENSEARCH_GO_DEBUG=true.

Build, test, and development

These variables are not read by the client at runtime. They configure the test harness and code-generation tooling, and are listed here so the reference is exhaustive. See DEVELOPER_GUIDE.md for the testing workflow.

Variable Used by Accepted values Default Meaning
OPENSEARCH_URL client + tests Comma-separated URL list unset Doubles as the test-cluster endpoint. (Also a runtime variable; see Connection.)
OPENSEARCH_VERSION test harness version string (e.g. 2.1.0, latest) unset OpenSearch server version under test.
SECURE_INTEGRATION test harness Bool true Run integration tests against a TLS + basic-auth cluster; false for an insecure cluster.
OPENSEARCH_NODE_COUNT test harness Integer 1 Expected node count for cluster-readiness checks.
OPENSEARCH_HEAP_SIZE test harness size string (e.g. 2g) 1g JVM heap per node for the test cluster.
TEST_PARALLEL test harness Integer CPU cores / 2 (min 1) Max parallel test functions (go test -parallel).
OPENSEARCH_GO_SKIP_JSON_COMPARE test harness presence (any value, including empty) unset When set, skips request/response JSON comparison. Detected by presence, not a parsed bool.
OSGEN_SKIP_GIT_CHECK cmd/osgen Bool false Bypass the git-root safety check during code generation.