Once your Horizon database is configured, you’re ready to run Horizon. To run Horizon you simply run
horizon serve, both of which start the HTTP server and start logging to standard out. When run, you should see output similar to:
INFO Starting horizon on :8000 pid=29013
The log line above announces that Horizon is ready to serve client requests. Note: the numbers shown above may be different for your installation. Next you can confirm that Horizon is responding correctly by loading the root resource. In the example above, that URL would be http://127.0.0.1:8000/, and simply running
curl http://127.0.0.1:8000/ shows you that the root resource can be loaded correctly.
If you didn’t set up a Stellar Core node yet, you may see an error like this:
ERRO[2019-05-06T16:21:14.126+08:00] Error getting core latest ledger err="get failed: pq: relation \"ledgerheaders\" does not exist"
Horizon provides most of its utility through ingested data. Your Horizon server can be configured to listen for and ingest transaction results from the connected Stellar Core instance.
To enable ingestion, you must either pass
--ingest=true on the command line or set the
INGEST environment variable to “true”. As of Horizon 1.0.0, you can start multiple ingesting machines in your cluster.
To enable ingestion of historical data from Stellar Core, you need to run
horizon db reingest range start end. If you’re running a full validator with published history archive, for example, you might want to ingest all of the network’s history. You can run this process in the background while your Horizon server is up. This continuously decrements the
history.elder_ledger in your /metrics endpoint until
NUM_LEDGERS is reached and the backfill is complete.
To reingest older ledgers — which you may need to do after a version upgrade — or to ingest ledgers closed by the network before you started Horizon use the
horizon db reingest range [START_LEDGER] [END_LEDGER] command:
horizon1> horizon db reingest range 1 10000 horizon2> horizon db reingest range 10001 20000 horizon3> horizon db reingest range 20001 30000 # ... etc.
This allows reingestion to be split up and done in parallel by multiple Horizon processes.
Over time, the recorded network history will grow unbounded, increasing storage used by the database. Horizon needs sufficient disk space to expand the data ingested from Stellar Core. Unless you need to maintain a history archive, you may configure Horizon to only retain a certain number of ledgers in the database. This is done using the
--history-retention-count flag or the
HISTORY_RETENTION_COUNT environment variable. Set the value to the number of recent ledgers you wish to keep around, and every hour the Horizon subsystem will reap expired data. Alternatively, you may execute the command
horizon db reap to force a collection.
Horizon tries to maintain a gap-free window into the history of the Stellar network. This reduces the number of edge cases that Horizon-dependent software must deal with in an attempt to make the integration process simpler. To maintain a gap-free history, Horizon needs access to all of the metadata produced by Stellar Core in the process of closing a ledger, and there are instances when this metadata can be lost. Usually, this loss of metadata occurs because the Stellar Core node went offline and performed a catchup operation when restarted.
To ensure that the metadata required by Horizon is maintained, you have several options: You may either set the
CATCHUP_COMPLETE Stellar Core configuration option to
true or configure
CATCHUP_RECENT to determine the amount of time your Stellar Core can be offline without having to rebuild your Horizon database.
Unless your node is a Full Validator which publishes an archive we do not recommend using the
CATCHUP_COMPLETE method, as this will force Stellar Core to apply every transaction from the beginning of the ledger, which will take an ever increasing amount of time. Instead, we recommend you set the
CATCHUP_RECENT config value. To do this, determine how long of a downtime you would like to survive (expressed in seconds) and divide by ten. This roughly equates to the number of ledgers that occur within your desired grace period since ledgers roughly close at a rate of one every ten seconds. With this value set, Stellar Core will replay transactions for ledgers that are recent enough, ensuring that the metadata needed by Horizon is present.
In the section above, we mentioned that Horizon tries to maintain a gap-free window. Unfortunately, it cannot directly control the state of stellar-core and so gaps may form due to extended down time. When a gap is encountered, Horizon will stop ingesting historical data and complain loudly in the log with error messages (log lines will include “ledger gap detected”). To resolve this situation, you must re-establish the expected state of the Stellar Core database and purge historical data from Horizon’s database. We leave the details of this process up to the reader as it is dependent upon your operating needs and configuration, but we offer one potential solution:
We recommend you configure the HISTORY_RETENTION_COUNT in Horizon to a value less than or equal to the configured value for CATCHUP_RECENT in Stellar Core. Given this situation, any downtime that would cause a ledger gap will require a downtime greater than the amount of historical data retained by Horizon. To re-establish continuity:
- Stop Horizon.
horizon db reapto clear the historical database.
- Clear the cursor for Horizon by running
stellar-core -c "dropcursor?id=HORIZON"(ensure capitilization is maintained).
- Clear ledger metadata from before the gap by running
stellar-core -c "maintenance?queue=true".
- Restart Horizon.
Endpoints that display state information are not available during initial state ingestion and will return a
503 Service Unavailable/
Still Ingesting error. An example is the
/paths endpoint (built using offers). Such endpoints will become available after state ingestion is done (usually within a couple of minutes).
State ingestion shouldn’t take more than a couple of minutes on an AWS
c5.xlarge instance, or equivalent.
It’s possible that the progress logs (see below) will not show anything new for a longer period of time or print a lot of progress entries every few seconds. This happens because of the way history archives are designed. The ingestion is still working but it’s processing entries of type
DEADENTRY’. If there is a lot of them in the bucket, there are no active entries to process. We plan to improve the progress logs to display actual percentage progress so it’s easier to estimate ETA.
If you see that ingestion is not proceeding for a very long period of time:
- Check the RAM usage on the machine. It’s possible that system ran out of RAM and it using swap memory that is extremely slow.
- If above is not the case, file a new issue in the Horizon repository.
This is by design. Horizon runs a state verifier routine that compares state in local storage to history archives every 64 ledgers to ensure data changes are applied correctly. If data corruption is detected, Horizon will block access to endpoints serving invalid data.
We recommend keeping this security feature turned on; however, if it’s causing problems (due to CPU usage) this can be disabled by
--ingest-disable-state-verification CLI param or
INGEST-DISABLE-STATE-VERIFICATION env variable.
If you were running the new system in the past during experimental stage (
ENABLE_EXPERIMENTAL_INGESTION flag) it’s possible that the old and new systems are not in sync. In such case, the upgrade code will activate and will make sure the data is in sync. When this happens you may see
Waiting for the next checkpoint... messages for up to 5 minutes.
In order to check the progress and the status of experimental ingestion you should check the logs. All logs connected to experimental ingestion are tagged with
It starts with informing you about state ingestion:
INFO[2019-08-29T13:04:13.473+02:00] Starting ingestion system from empty state... pid=5965 service=expingest temp_set="*io.MemoryTempSet" INFO[2019-08-29T13:04:15.263+02:00] Reading from History Archive Snapshot ledger=25565887 pid=5965 service=expingest
During state ingestion, Horizon will log number of processed entries every 100,000 entries (there are currently around 7M entries in the public network):
INFO[2019-08-29T13:04:34.652+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=100000 pid=5965 service=expingest INFO[2019-08-29T13:04:38.487+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=200000 pid=5965 service=expingest INFO[2019-08-29T13:04:41.322+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=300000 pid=5965 service=expingest INFO[2019-08-29T13:04:48.429+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=400000 pid=5965 service=expingest INFO[2019-08-29T13:05:00.306+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=500000 pid=5965 service=expingest
When state ingestion is finished, it will proceed to ledger ingestion starting from the next ledger after checkpoint ledger (25565887+1 in this example) to update the state using transaction meta:
INFO[2019-08-29T13:39:41.590+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=5300000 pid=5965 service=expingest INFO[2019-08-29T13:39:44.518+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=5400000 pid=5965 service=expingest INFO[2019-08-29T13:39:47.488+02:00] Processing entries from History Archive Snapshot ledger=25565887 numEntries=5500000 pid=5965 service=expingest INFO[2019-08-29T13:40:00.670+02:00] Processed ledger ledger=25565887 pid=5965 service=expingest type=state_pipeline INFO[2019-08-29T13:40:00.670+02:00] Finished processing History Archive Snapshot duration=2145.337575904 ledger=25565887 numEntries=5529931 pid=5965 service=expingest shutdown=false INFO[2019-08-29T13:40:00.693+02:00] Reading new ledger ledger=25565888 pid=5965 service=expingest INFO[2019-08-29T13:40:00.694+02:00] Processing ledger ledger=25565888 pid=5965 service=expingest type=ledger_pipeline updating_database=true INFO[2019-08-29T13:40:00.779+02:00] Processed ledger ledger=25565888 pid=5965 service=expingest type=ledger_pipeline INFO[2019-08-29T13:40:00.779+02:00] Finished processing ledger duration=0.086024492 ledger=25565888 pid=5965 service=expingest shutdown=false transactions=14 INFO[2019-08-29T13:40:00.815+02:00] Reading new ledger ledger=25565889 pid=5965 service=expingest INFO[2019-08-29T13:40:00.816+02:00] Processing ledger ledger=25565889 pid=5965 service=expingest type=ledger_pipeline updating_database=true INFO[2019-08-29T13:40:00.881+02:00] Processed ledger ledger=25565889 pid=5965 service=expingest type=ledger_pipeline INFO[2019-08-29T13:40:00.881+02:00] Finished processing ledger duration=0.06619956 ledger=25565889 pid=5965 service=expingest shutdown=false transactions=29 INFO[2019-08-29T13:40:00.901+02:00] Reading new ledger ledger=25565890 pid=5965 service=expingest INFO[2019-08-29T13:40:00.902+02:00] Processing ledger ledger=25565890 pid=5965 service=expingest type=ledger_pipeline updating_database=true INFO[2019-08-29T13:40:00.972+02:00] Processed ledger ledger=25565890 pid=5965 service=expingest type=ledger_pipeline INFO[2019-08-29T13:40:00.972+02:00] Finished processing ledger duration=0.071039012 ledger=25565890 pid=5965 service=expingest shutdown=false transactions=20
Horizon ingests ledger data from a connected instance of Stellar Core. In the event that Stellar Core stops running (or if Horizon stops ingesting data for any other reason), the view provided by Horizon will start to lag behind reality. For simpler applications, this may be fine, but in many cases this lag is unacceptable and the application should not continue operating until the lag is resolved.
To help applications that cannot tolerate lag, Horizon provides a configurable “staleness” threshold. Given that enough lag has accumulated to surpass this threshold (expressed in number of ledgers), Horizon will only respond with an error:
stale_history. To configure this option, use either the
--history-stale-threshold command line flag or the
HISTORY_STALE_THRESHOLD environment variable. NOTE: non-historical requests (such as submitting transactions or finding payment paths) will not error out when the staleness threshold is surpassed.
Last updated Apr. 12, 2021