Instead of trying to decode and re-encode status logs, we now write them directly as they come in.
This change prevents future changes to the osquery status log file format (addition and deletion of fields ) from
affecting Fleet. A similar change was implemented in #1636 for result logs.
Closes#1664
Initially fleet decoded the incoming JSON sent to the log endpoint.
Then the log event would be written to a log writer by calling json.Encoder{}.Encode.
Re-encoding logs is lossy; whenever a new field is sent by osqueryd we don't keep up with them.
Instead of caring about the content of the OsqueryResultLog, fleet will now write all log results
exactly as sent to the server by osqueryd.
Closes#1632Closes#1615
This PR adds support for file integrity monitoring. This is done by providing a simplified API that can be used to PATCH/GET FIM configurations. There is also code to build the FIM configuration to send back to osquery. Each PATCH request, if successful, replaces Fleet's existing FIM configuration. For example:
curl -X "PATCH" "https://localhost:8080/api/v1/kolide/fim" \
-H "Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzZXNzaW9uX2tleSI6IkVhaFhvZWswMGtWSEdaTTNCWndIMnhpYWxkNWZpcVFDR2hEcW1HK2UySmRNOGVFVE1DeTNTaUlFWmhZNUxhdW1ueFZDV2JiR1Bwdm5TKzdyK3NJUzNnPT0ifQ.SDCHAUA1vTuWGjXtcQds2GZLM27HAAiOUhR4WvgvTNY" \
-H "Content-Type: application/json; charset=utf-8" \
-d $'{
"interval": 500,
"file_paths": {
"etc": [
"/etc/%%"
],
"users": [
"/Users/%/Library/%%",
"/Users/%/Documents/%%"
],
"usr": [
"/usr/bin/%%"
]
}
}'
Fix for issue where osquery sends empty strings where we expect integers in detail queries. We handle empty strings in these cases by changing them to "0" and then letting the different conversion functions change the "0" string into the appropriate integer type. This has been tested against running osquery hosts.
Closes#1521
We now track the `config_tls_refresh`, `distributed_interval` and
`logger_tls_period` flag values for each host. Each value is updated by a
detail query agains the `osquery_flags` table, because they may be specified
outside of Kolide. The flags that can be specified within Kolide are also
updated when a config is returned to the host that changes their value.
This will enable us to do a more accurate per-host online status calculation as
discussed in #1419.
Made log rotation for osquery results and status logs optional. This required writing the logwriter package which is a drop in replacement for lumberjack. We still use lumberjack if the log rotation flag --osquery_enable_log_rotation flag is set. Note that the performance of the default is quite a bit better than lumberjack.
BenchmarkLogger-8 2000000 747 ns/op
BenchmarkLumberjack-8 1000000 1965 ns/op
PASS
BenchmarkLogger-8 2000000 731 ns/op
BenchmarkLumberjack-8 1000000 2040 ns/op
PASS
BenchmarkLogger-8 2000000 741 ns/op
BenchmarkLumberjack-8 1000000 1970 ns/op
PASS
BenchmarkLogger-8 2000000 737 ns/op
BenchmarkLumberjack-8 1000000 1930 ns/op
PASS
The seen time should only be updated once per request from the osquery agent to
the Kolide server. We now do that only in AuthenticateHost (which every request
besides enrollment must go through).
Return `accelerate: 10` with distributed queries if we do not have host
details. This facilitates the host quickly joining all expected labels, as
`platform` gated label queries will not be returned until the detail queries
return with the platform.
Fixes#1421.
On CentOS6 there is a bug in which osquery incorrectly reports an empty string
for platform. This PR fixes our detection of centos in this case.
Fixes#1339
Notable refactoring:
- Use stdlib "context" in place of "golang.org/x/net/context"
- Go-kit no longer wraps errors, so we remove the unwrap in transport_error.go
- Use MakeHandler when setting up endpoint tests (fixes test bug caught during
this refactoring)
Closes#1411.
These decorators were removed in #953 due to an osquery bug. That bug is now
fixed, and we are adding the decorators back. We also now use `load` decorators
rather than `interval` decorators because they seem to function more reliably.
Previously we were using `build_platform`, which does not always properly
reflect the platform of the host running osquery. Now we should properly
retrieve the platform.
Fixes#1264
* If a detail query doesn't return results, log problem but don't interrupt additional query processing
* changed arg to ingest func to logger interface
Saving a new detail update time when the host details were not actually updated
caused detail updates to be missed. This PR fixes the existing test to catch
the bug, and fixes the bug.
As of recently, osquery will report when a distributed query fails. We now
expose errors over the results websocket. When a query errored on the host, the
`error` key in the result will be non-null. Note that osquery currently doesn't
provide any details so the error string will always be "failed". I anticipate
that we will fix this and the string is included for future-proofing.
Successful result:
```
{
"type": "result",
"data": {
"distributed_query_execution_id": 15,
"host": {
... omitted ...
},
"rows": [
{
"hour": "1"
}
],
"error": null
}
}
```
Failed result:
```
{
"type": "result",
"data": {
"distributed_query_execution_id": 14,
"host": {
... omitted ...
},
"rows": [
],
"error": "failed"
}
}
```
A distributed query campaign can be "orphaned" (left in the QueryRunning state)
if the Kolide server restarts while it is running, or other weirdness occurs.
When this happens, no subscribers are waiting to read results written by
osqueryd agents, but the agents continue to receive the query. Previously, this
would cause us to error on ingestion.
The new behavior will instead set the campaign to completed when it detects
that it is orphaned. This should prevent sending queries for which there is no
subscriber.
- New NoSubscriber error interface in pubsub
- Detect NoSubscriber errors and close campaigns
- Tests on pubsub and service methods
Fixes#695
* Moving query attributes from the query object to the pack-query relationship
* some additional tests
* http request parsing test
* QueryOptions in new test_util code
* initial scaffolding of new request structures
* service and datastore
* test outline
* l2 merge conflict scrub
* service tests for scheduled query service
* service and datastore tests
* most endpoints and transports
* order of values are not deterministic with inmem
* transport tests
* rename PackQuery to ScheduledQuery
* removing existing implementation of adding queries to packs
* accounting for the new argument to NewQuery
* fix alignment in sql query
* removing underscore
* add removed to the datastore
* removed differential from the schema
Removed Gorm, replaced it with Sqlx
* Added SQL bundling command to Makfile
* Using go-kit logger
* Added soft delete capability
* Changed SearchLabel to accept a variadic param for optional omit list
instead of array
* Gorm removed
* Refactor table structures to use CURRENT_TIMESTAMP mysql function
* Moved Inmem datastore into it's own package
* Updated README
* Implemented code review suggestions from @zwass
* Removed reference to Gorm from glide.yaml
- Introduce a new pattern for defining/ingesting detail queries
- Add many relevant host details:
- Platform
- osquery Version
- Memory
- Hostname
- UUID
- OS Version
- Uptime
- Primary interface MAC
- Primary interface IP
- Fix parsing for inconsistent JSON schema returned from osquery
- Tests
- Establish a pattern for host authentication
- Establish a pattern for error JSON
- Add transport and make endpoint functions
- Fix discovered bugs + update tests
