Skip to content

doc(server): sync store with master - #479

Open
bitflicker64 wants to merge 1 commit into
apache:masterfrom
bitflicker64:doc-sync/server-store
Open

doc(server): sync store with master#479
bitflicker64 wants to merge 1 commit into
apache:masterfrom
bitflicker64:doc-sync/server-store

Conversation

@bitflicker64

@bitflicker64 bitflicker64 commented Sep 5, 2026

Copy link
Copy Markdown
Contributor

Syncs the HugeGraph-Store quickstart page with apache/hugegraph master (commit 36811483a). Both content/en and content/cn are updated.

page what was wrong what changed source (file:line on master)
quickstart/hugegraph/hugegraph-hstore.md The application.yml block did not match the shipped template: pdserver.address was shown as 127.0.0.1:8686 and the whole management block was missing Block now mirrors the shipped file byte for byte, pdserver.address: localhost:8686 plus the management section hugegraph-store/hg-store-dist/src/assembly/static/conf/application.yml:18-63
quickstart/hugegraph/hugegraph-hstore.md application-pd.yml and its rocksdb.* keys were not documented at all, even though spring.profiles.include: pd pulls the file in New section 4.2 with the file contents, and a RocksDB table for total_memory_size, write_buffer_size, min_write_buffer_number_to_merge, write_buffer_ratio hugegraph-store/hg-store-dist/src/assembly/static/conf/application-pd.yml:18-34; hugegraph-store/hg-store-rocksdb/src/main/java/org/apache/hugegraph/rocksdb/access/RocksDBOptions.java:36-53,220-240; hugegraph-store/hg-store-core/src/main/java/org/apache/hugegraph/store/options/RaftRocksdbOptions.java:144-153
quickstart/hugegraph/hugegraph-hstore.md No defaults or meanings for any config key, and most keys the node reads were missing (app.fake-pd, app.raft-path, app.placeholder-size, app.label.*, grpc.server.wait-time, ten raft.* keys, ave-logEntry-size-ratio, thread.pool.*, query.push-down.*, job.*, fake-pd.*, arthas.*) New section 4.3 with per-prefix tables giving each key its template value, code default and meaning hugegraph-store/hg-store-node/src/main/java/org/apache/hugegraph/store/node/AppConfig.java:40-313; hugegraph-store/hg-store-node/src/main/java/org/apache/hugegraph/store/node/listener/PlaceHolderListener.java:41-68; hugegraph-store/hg-store-node/src/main/java/org/apache/hugegraph/store/node/task/TTLCleaner.java:107-113,279
quickstart/hugegraph/hugegraph-hstore.md Only the -d start flag was documented Flag table for -d, -g, -j, -y, including that -g g1 actually aborts the start even though the script's usage line suggests it, plus the real default heap (-Xms512m, -Xmx between 512 MB and 2048 MB), the ulimit preflight and jemalloc preload hugegraph-store/hg-store-dist/src/assembly/static/bin/start-hugegraph-store.sh:45-176; hugegraph-store/hg-store-dist/src/assembly/static/bin/util.sh:179-197
quickstart/hugegraph/hugegraph-hstore.md restart-hugegraph-store.sh was not mentioned, and the stop script's behavior was not described New section 5.3 for restart, and the stop section now explains the bin/pid file and the 30 second shutdown wait hugegraph-store/hg-store-dist/src/assembly/static/bin/restart-hugegraph-store.sh:30-32; hugegraph-store/hg-store-dist/src/assembly/static/bin/stop-hugegraph-store.sh:35-48
quickstart/hugegraph/hugegraph-hstore.md Startup order relative to PD and Server was a single sentence, and the pd.initial-store-list requirement was missing New section 5.4: PD, then Store, then Server, plus the fact that a Store absent from pd.initial-store-list registers as Pending and never reaches Up, and that Store retries registration when PD is unreachable hugegraph-pd/hg-pd-core/src/main/java/org/apache/hugegraph/pd/StoreNodeService.java:161-167,218-220; hugegraph-store/hg-store-core/src/main/java/org/apache/hugegraph/store/HeartbeatService.java:123-160; docker/docker-compose-3pd-3store-3server.yml:46-60,74-90
quickstart/hugegraph/hugegraph-hstore.md /actuator/health was the only endpoint listed, and /v1/health was described only in the Docker section New section 7.1 listing the endpoints that exist in hg-store-node: /v1/health, /actuator/health, /actuator/prometheus, /, /-/state, /-/echo, /-/scan, /v1/partitions, /v1/partition/{id}, /metrics/system, /metrics/drive, /metrics/raft, with a note on the maintenance endpoints hugegraph-store/hg-store-node/src/main/java/org/apache/hugegraph/store/node/controller/HgStoreStatusController.java:49-117; .../IndexAPI.java:48-54; .../PartitionAPI.java:66-205; .../HgStoreMetricsController.java:46-62; .../RaftAPI.java:42-45; .../FixGraphIdController.java:64-479
quickstart/hugegraph/hugegraph-hstore.md Docker section listed only docker-compose-3pd-3store-3server.yml and called it a 3 node cluster Added docker-compose-hstore.yml (1 PD + 1 Store + 1 Server + 1 Hubble) and corrected both topologies to include Hubble docker/docker-compose-hstore.yml:53-103; docker/docker-compose-3pd-3store-3server.yml:96-231; docker/README.md:16-17
quickstart/hugegraph/hugegraph-hstore.md The env var table did not say which config key each variable sets, and the image behavior was undocumented Added a "Maps to" column, plus the SPRING_APPLICATION_JSON mechanism, JAVA_OPTS default, STDOUT_MODE=true, the HEALTHCHECK parameters and the fact that only 8520 is declared EXPOSE hugegraph-store/hg-store-dist/docker/docker-entrypoint.sh:45-80; hugegraph-store/Dockerfile:42-69
quickstart/hugegraph/hugegraph-hstore.md Docker cluster snippet did not explain that only the host ports differ between store0/1/2, and did not mention the matching PD setting Snippet now labels the published host ports per node and notes HG_PD_INITIAL_STORE_LIST: store0:8500,store1:8500,store2:8500 docker/docker-compose-3pd-3store-3server.yml:151-194,107
quickstart/hugegraph/hugegraph-hstore.md Building Store on its own was not covered Added the hugegraph-struct first step and the hg-store-dist package command, plus what the assembled directory contains hugegraph-store/hg-store-dist/src/assembly/descriptor/server-assembly.xml:27-60; hugegraph-store/AGENTS.md:62-67,187
quickstart/hugegraph/hugegraph-hstore.md (cn only) The tar extraction step used apache-hugegraph-hstore-incubating-1.7.0, which is not the directory the release produces Corrected to apache-hugegraph-store-incubating-1.7.0, matching the en page hugegraph-store/pom.xml:48; install-dist/pom.xml:54
quickstart/hugegraph/hugegraph-hstore.md Overview did not say how storage and replication actually work Added one paragraph: RocksDB per node, one Raft group per partition, PD assigns partitions and Store reports back over heartbeat, Server reaches Store over gRPC hugegraph-store/README.md:10,77-79; hugegraph-store/hg-store-core/src/main/java/org/apache/hugegraph/store/options/HgStoreEngineOptions.java:34-72
quickstart/hugegraph/hugegraph-hstore.md Em dashes were used as the "no default" marker and in the daemon mode note Replaced with n/a and plain punctuation n/a
quickstart/hugegraph/hugegraph-hstore.md The PD registration check ran curl http://localhost:8620/v1/stores with no credentials, which PD rejects with {"status":-1,"error":"Unauthorized!"} The call now sends -u store:admin, with a note that the user name must be one of hg, store, hubble, vermeer, that the password is not validated yet, and that only /v1/health, /actuator/* and /v1/prom/targets/* skip the interceptor hugegraph-pd/hg-pd-service/src/main/java/org/apache/hugegraph/pd/rest/interceptor/AuthenticationConfigurer.java:33-35; hugegraph-pd/hg-pd-service/src/main/java/org/apache/hugegraph/pd/rest/interceptor/RestAuthentication.java:48-53; hugegraph-pd/hg-pd-service/src/main/java/org/apache/hugegraph/pd/service/interceptor/Authentication.java:62,79-87

Rewrote the HugeGraph-Store quickstart to match hugegraph master: the
application.yml block now mirrors the shipped template, application-pd.yml
and its rocksdb keys are documented, and a full configuration reference
lists every key the node reads with its template value and code default.
Also documented the -g/-j/-y start flags, the restart script, the PD to
Store to Server startup order and the pd.initial-store-list requirement,
the Store REST endpoints that actually exist in hg-store-node, the minimal
docker-compose-hstore.yml topology, and the image defaults. Fixed the
wrong Chinese tar directory name.

The PD /v1/stores check now passes basic auth credentials, since PD's REST
interceptor covers every path except /v1/health, /actuator/* and
/v1/prom/targets/*.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant