Bug 1879837
| Summary: | DonNew audit profiles AllRequestBodies and WriteRequestBodies generate log entries that are too big | ||||||
|---|---|---|---|---|---|---|---|
| Product: | OpenShift Container Platform | Reporter: | Frederic Giloux <fgiloux> | ||||
| Component: | kube-apiserver | Assignee: | Abu Kashem <akashem> | ||||
| Status: | CLOSED ERRATA | QA Contact: | Ke Wang <kewang> | ||||
| Severity: | medium | Docs Contact: | |||||
| Priority: | medium | ||||||
| Version: | 4.6 | CC: | aos-bugs, jhou, mfojtik, sttts, xxia | ||||
| Target Milestone: | --- | ||||||
| Target Release: | 4.6.0 | ||||||
| Hardware: | Unspecified | ||||||
| OS: | Unspecified | ||||||
| Whiteboard: | |||||||
| Fixed In Version: | Doc Type: | If docs needed, set a value | |||||
| Doc Text: | Story Points: | --- | |||||
| Clone Of: | Environment: | ||||||
| Last Closed: | 2020-10-27 16:41:53 UTC | Type: | Bug | ||||
| Regression: | --- | Mount Type: | --- | ||||
| Documentation: | --- | CRM: | |||||
| Verified Versions: | Category: | --- | |||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||
| Embargoed: | |||||||
| Attachments: |
|
||||||
Managed fields sounds legit. But other than that, what do you expect when telling the system to log everything? I tend to close this as NOTABUG. It works exactly as designed. Hi Stefan. As you know "managedFields" are a new addition (Kubernetes 1.18), which was not available in 3.x. Would you not agree that they don't bring anything to audit logs? I believe audit profiles intent is to make audit customization possible in 4.x as it was in 3.x. Now due to the point reported I am not sure that the new profiles could be used in production clusters, which may question the usefulness of the feature. If you have evidences of the opposite feel free to close as NOTABUG or change the bugzilla to an RFE. No, I agree that the value of managed fields is limited and we should probably not log them. We will look into the managedFields topic. This sounds easy to do. Great! I have discussed a strategy with Stefan. An upstream fix is in the works. Moving it to 4.7 now. If we can get the upstream fix done in time then we will take a stab t carrying/porting it to 4.6. We have opened a PR upstream to discard the managed fields - https://github.com/kubernetes/kubernetes/pull/94986. If it lands in time we may have a chance to get it to 4.6. Very good. Fingers crossed. moving it to 4.6 since we are carrying the upstream PR - https://github.com/openshift/kubernetes/pull/375 $ oc get clusterversion
NAME VERSION AVAILABLE PROGRESSING SINCE STATUS
version 4.6.0-0.nightly-2020-09-26-202331 True False 3h56m Cluster version is 4.6.0-0.nightly-2020-09-26-202331
$ oc get apiserver/cluster -ojson | jq .spec.audit
{
"profile": "AllRequestBodies"
}
$ oc debug node/kewang27dr1-btdxn-m-0.c.openshift-qe.internal
...
sh-4.4# chroot /host
sh-4.4# cd /var/log/kube-apiserver/
sh-4.4# jq . audit.log > audit.json
sh-4.4# ls
audit-2020-09-27T08-53-40.952.log audit-2020-09-27T09-03-49.795.log audit-2020-09-27T09-14-52.678.log audit-2020-09-27T09-25-26.290.log termination.log
audit-2020-09-27T08-57-18.934.log audit-2020-09-27T09-07-27.321.log audit-2020-09-27T09-18-00.608.log audit.json
audit-2020-09-27T09-00-50.873.log audit-2020-09-27T09-11-14.594.log audit-2020-09-27T09-21-50.698.log audit.log
sh-4.4# cat audit.json | grep managedFields
"managedFields": null,
"managedFields": null,
"managedFields": null,
The ManagedFields of request and response bodies are discarded in audit logs as expected, so move the bug verified.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (OpenShift Container Platform 4.6 GA Images), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2020:4196 |
Created attachment 1715176 [details] Request record in the audit logs Description of problem: The new profiles AllRequestBodies and WriteRequestBodies store an awful amount of data in the audit logs. For a single request listing pods and returning ~30 instances: $ oc get pods -n openshift-kube-apiserver A json content of 296KB was stored in the logs. As list commands happen all the time in a cluster (reconciliation loops) and thousands of requests per second may be processed in big clusters under load this may be a significant amount of storage that gets consumed very quickly. SIEM Consumers may also not be able to cope. As some of the data stored is not useful it could be trimmed out, for instance: - Managed fields (most of the payload) - Column definitions and descriptions Version-Release number of selected component (if applicable): 4.6.0-fc.5 How reproducible: Always Steps to Reproduce: 1. Patch the cluster configuration: $ oc patch apiservers.config.openshift.io cluster --type='json' -p='[{"op": "replace", "path": "/spec/audit/profile", "value":"AllRequestBodies"}]' 2. Send a test query: $ date & oc get pods -n openshift-kube-apiserver 3. Retrieve the query in the logs. I used: $ oc adm node-logs ci-ln-4hgkjxk-f76d1-nrx5r-master-0 --path=kube-apiserver/audit-2020-09-16T16-01-24.053.log | grep "openshift-kube-apiserver.*pods.*list.*system\:admin" Actual results: See attachment Expected results: Slimmer result. At least without the "managed fields". Additional info: