> For the complete documentation index, see [llms.txt](https://docs.prismacloud.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.prismacloud.io/release-notes/look-ahead-planned-updates-on-prisma-cloud/look-ahead-secure-the-runtime.md). # Look Ahead Updates to Secure the Runtime Here are the changes planned in the next Prisma Cloud release to ensure the security of your runtime. Read this section to learn about what is planned in the upcoming 34.05 release. The Look Ahead announcements are for an upcoming release and is not a cumulative list of all announcements. The details and functionalities listed below are a preview and the actual release date is subject to change. * [Enhancements](#enhancements) * [Changes in Behavior](#changes-in-existing-behavior) * [Known and Fixed Issues](#known-and-fixed-issues) ### Enhancements | Enhancement | Details | | ----------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Support for Rocky Linux 10.x (Red Quartz) | Prisma Cloud Compute now supports Rocky Linux 10.x (Red Quartz) for vulnerability and compliance scanning and for runtime protection. | | Rocky Linux — source-package and epoch support | During scanning, Rocky Linux binary packages are now linked to their source package (for example, `expat-devel` and `expat-libs` linked to `expat`). This ensures that CVE matches against the Rocky Linux errata feed are no longer missed. Version comparison also honors the RPM epoch, which eliminates a class of false positives on Rocky Linux images. | | Support for RKE2 v1.35 | Defenders now deploy successfully on RKE2 v1.35 clusters. Previously, defenders failed to connect to Console because CRI-client initialization could not determine the container runtime version on newer RKE2 releases. | | Host defender support on RHEL 10 with nftables | Host defenders deployed on RHEL 10 with the `-n` (nftables) install flag no longer log spurious `iptables: executable file not found in $PATH` errors while evaluating the Linux CIS firewall configuration. | | Generic distro-based CVE exclusion via custom feed entries | Custom CVE exclusions now support any OS distribution — Windows, Ubuntu, Red Hat, Alpine, and others — extending the existing PAN-OS / GKE mechanism. Administrators can upload custom `excludedCve` entries per target distribution through the portal to suppress specific CVEs from vulnerability results. | | False-positive suppression for CVEs on OpenShift nodes | OpenShift node scans now filter out CVEs that Red Hat has marked as "not affected" for the OpenShift version in the VEX feed. This eliminates false positives such as `CVE-2025-30204` on OpenShift 4.14 RHCOS nodes. | | Environment variables printed on defender startup | On startup, the defender now logs all of its environment variables, making support cases involving env-var configuration significantly faster to triage. | | Defender image — Red Hat Ecosystem Catalog certification | The defender image now includes the required labels (`name`, `vendor`, `version`, `release`, `summary`, `description`, `maintainer`) and a `/licenses` folder with license files (MIT, Apache, and others), enabling certification in the Red Hat Ecosystem Catalog. | | Agentless OCI — compartment ID support | The Agentless scanner for Oracle Cloud Infrastructure now supports specifying a compartment ID, which allows customers to scope agentless scans to a specific OCI compartment. | | Agentless AWS — expanded fallback instance types | The Agentless scanner now uses additional fallback instance types beyond `m5.2xlarge`, `m4.2xlarge`, `m3.2xlarge`, `t2.2xlarge`, and `m6i.2xlarge`. This allows agentless scans to succeed in AWS regions where the current defaults are unavailable (for example, `af-south-1`, `me-central-1`, `il-central-1`, `ap-southeast-4`, `eu-central-2`, and others). | | Registry scan — image deduplication by digest | Registry scans configured with a Cap now deduplicate images by digest before enforcing the Cap, so the same image referenced by multiple tags is no longer scanned repeatedly or double-counted against the Cap. | | Vulnerability / Compliance Explorer — daily refresh at very large scale | The daily compliance-stats aggregation in `db.AggregatedComplianceData` is restructured so it no longer accumulates a `$push` array that exceeds MongoDB's hard-coded 100 MiB internal limit. This unblocks daily refresh of the Vulnerability and Compliance Explorer for very large tenants (\~4M+ container compliance hits). A workaround is also available: set the new `MONGO_DB_CUSTOM_CONFIG_PATH` environment variable to mount a custom mongodb configuration that raises `internalQueryMaxPushBytes` from 100 MiB to 500 MiB. | | LDAP users in multiple groups — access to all assigned collections | On on-prem consoles using LDAP, when a user belongs to multiple LDAP groups mapped to custom roles, the user is now granted access to every collection associated with those groups instead of only the first collection. | ### Changes in Existing Behavior | Feature | Description | | ------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Removal of environment-variable printing from defender failure log | The defender no longer prints its environment variables to the log on failure. This complements the new startup-time env-var dump (see Enhancements) and prevents environment state from appearing in failure logs. | ### Known and Fixed Issues #### Fixed Issues | Issue | Details | | ---------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | PSIRT — Access Control on Delete Defender API | The `DELETE /api/v1/defenders/{id}` endpoint now validates that the target defender belongs to the caller's assigned account group, which prevents access keys with Defender Management role and Account-Group restrictions from deleting out-of-scope defenders. See [pan.dev](https://pan.dev/prisma-cloud/api/cwpp/delete-defenders-id/). | | API Key bypassed UI-login restriction | Access keys are no longer able to authenticate to the Console UI when the `core-block-user-direct-login-enabled` policy is intended to block them. Validation of external tokens (`saasAuthenticator.Authenticate` → `authenticateToken`) strips the `accessUI` permission bit for access-key logins. | | CNNF \| rules set even if the feature is disabled | Regression from CWP-64459 (Quinn Update 4) that caused Twistlock iptables rules (`TWISTLOCK-NET-PRE`, `TWISTLOCK-NET-POST`) to be programmed even when CNNF was disabled. The firewall update path is now gated so no rules are added when CNNF is off. | | WAAS — defender memory forced to 4 GB after rule enabled | After enabling a WAAS rule on a cluster, the defender's max memory limit was forced to 4 GB. This is fixed and the defender no longer decreases memory below an already-configured value. | | App-embedded \| defender failed to run httpd with FILESYSTEM\_MONITORING turned on | The app-embedded defender now runs `httpd`-based applications correctly when the `FILESYSTEM_MONITORING` capability is enabled. Previously the combination caused httpd to fail to start. | | Container defender — unhealthy state due to missing awk | On RHEL 9 hosts running the container defender, `defender_health_check.sh` failed with `awk: command not found` starting in Quinn Update 4 (34.04.145/146), causing the container to report unhealthy. The health check no longer depends on `awk`. | | Defender CRI client initialization stuck in infinite loop after node reboot | On CRI-type environments (containerd / CRI-O), if the defender started before the container runtime finished initializing (typical after a node reboot on RKE2), it would loop forever with `failed to initialize CRI client: runtime version cannot be determined`. CRI-socket discovery now re-runs on every retry, so the defender recovers automatically once the runtime is available. | | Defender \| Windows/containerd \| getImageByName fails for multi-arch images | Windows containerd manifest-list resolution now matches on `major.minor.build` OS version (and falls back to the closest Windows `arch+os` entry) instead of requiring an exact `OSVersion` prefix match. This resolves `failed to resolve image mcr.microsoft.com/oss/v2/kubernetes/pause:3.10.1` on Windows Server 2022 (build 20348) nodes. | | Defender \| Windows/containerd \| One unresolvable image emptied Compliance > Containers | On Windows containerd nodes, a single container whose image failed to resolve caused `ListContainers` to abort and report zero containers everywhere in the UI. The defender now logs and skips the offending container, so the rest of the inventory is reported. | | AKS windows defender — error status and partial results | On AKS Windows Server 2022 nodes, the defender showed an error status, and only a subset of running containers appeared in runtime models, image scan results, and compliance. This was caused by the same image-resolution issue as CWP-64904 and CWP-64905 and is fixed alongside those tickets. | | runc \| replaceProxyRuncBinary \| runc.tw may be corrupted | When creating `/var/lib/twistlock/bin/runc.tw`, the defender now uses a "safe copy" pattern (write to a temp path, then rename) and skips the copy entirely when the binary is unchanged (same hash and size). This prevents a corrupt `runc.tw` from being left behind if the copy is interrupted during startup or a Block-policy switch. | | FP on RHEL EUS images — EUS content sets dropped by findRhelContentSets() | On RHEL EUS images with multiple content-set manifests, `findRhelContentSets()` returned an empty set because of a strict `>` comparison on `image_layer_index` when both manifests had index `0`. As a result, fixes shipped via EUS advisories (for example, `pam-1.5.1-24.el9_4.1` from `RHSA-2025:15102`) were reported as unfixed against the non-EUS fix version. The comparison is corrected, and the EUS content set is now honored. | | Intelligence — CVE description DB insert failures | The Intelligence pipeline no longer fails to insert CVE descriptions into the database, so descriptions in the Console reflect the latest feed content. | | Duplicate jar/go vulnerabilities when package keys collide | Fixed a regression from CWP-64606 (Quinn Update 4) in which jar and Go CVEs mapped to multiple package keys were no longer deduplicated at load time, producing duplicate CVE rows in scan results. Load-time deduplication is restored while the CWP-64606 performance gains are preserved. | | Intelligence builder — repeated CVE-validation logs suppressed | The Intelligence builder now suppresses repeated CVE-validation log entries, so recurring vendor-feed issues no longer flood the log. | | Intelligence \| staging \| on-prem — unexpected error connecting to IS | Fixed the on-prem Intelligence Stream error: `unexpected intelligence error connection to Intelligence Stream failed` returned by `/api/v1/statuses/intelligence` (with `connected: false`) after a manual auxiliary-identifiers upload. | | Agentless-AWS — on-demand fallback on spot capacity errors | Agentless AWS now correctly falls back to on-demand instances when AWS returns `InsufficientInstanceCapacity` for spot requests, not only when it returns `MaxSpotInstanceCountExceeded` or "no spot capacity". This aligns behavior with the [Agentless documentation](https://docs.prismacloud.io/en/compute-edition/34/admin-guide/agentless-scanning/agentless-scanning#csp-cost). | | Agentless — scans stuck in loop after subscription disabled | If the Console was restarted mid-scan and the customer then disabled the Agentless subscription, account statuses became stuck in `cleanup` and the UI continued to show an ongoing scan. Account statuses are now reset after the post-restart cleanup when no new scan is initiated. | | Misconfigured cloud accounts causing scan halt | Misconfigured cloud accounts in the Agentless orchestrator no longer cause the entire scan cycle to halt. | | Defender — Python 3.10+ version detection for dynamically loaded interpreters | The defender now correctly detects Python 3.10.x and later versions when the interpreter is dynamically loaded from `libpython3.10.so.1.0` (or similar), instead of misreporting version `1.2.8` from an unrelated `libpython` string match. | | Red Hat VEX — CVEs with "known\_affected" and "Fix deferred" reported correctly | Red Hat CVEs derived from a `known_affected` product with `none_available` category and "Fix deferred" details are now reported with status "fix deferred" instead of "affected". The conditions field is preserved as `*`. | | Flaky TestServerDefender — nil pointer dereference in socketConn.close() during upgrade | Fixed a nil-pointer dereference in `socketConn.close()` during defender upgrade (when fsmon had not yet connected to the event socket) by adding an explicit `UnixConn` nil check. | | controller\_filesystem \| Healthcheck isn't set for process events | Process events fired by the filesystem controller now populate the `Healthcheck` field on `ProcNotifyEvent`, so container Healthcheck processes are not blocked by a runtime Prevent rule. | | Vulnerabilities \| Remove libnghttp2 | The unused `libnghttp2` package is removed from the defender base image to eliminate a source of recurring vulnerabilities. The defender uses Go's HTTP stack directly. | | Address CVEs in PCC | Console vulnerability cleanup — upstream modules are upgraded to remediate outstanding CVEs in the PCC image. | | Daemonset — Warning in defender yaml about deprecated "apparmor" | The daemonset YAML no longer emits the `spec.template.metadata.annotations[container.apparmor.security.beta.kubernetes.io/twistlock-defender]: deprecated since v1.30; use the "appArmorProfile" field instead` warning when applied on Kubernetes v1.30+. | #### Known Issues | Issue | Details | | -------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Ava — Image scan opens binary for write and blocks other processes | During image scanning, AVA may open a binary for write and hold the file open, which causes other processes (for example, LTEE) to fail launching with `Text file busy`. Under investigation. | | QA validation for CWP-64703 | The Defender serverless Java package-dependency change from CWP-64703 was merged to master and Update 4 before formal QA sign-off. Full QA validation is scheduled. | | Rhel10 — Some containers get "base layer OS not supported" in Agentless scanning | Agentless scanning of certain RHEL 10 UBI-based images (for example, `registry.redhat.io/rhel10/rtla:10.0-1760386166`, `registry.redhat.io/ubi10/ubi:10.0-1755007654`) fails with "base layer OS not supported", although defender-based scans of the same images succeed. Fix targeted for a later release. | | VM image scan — Error when trying to scan VM images in Azure Gov Cloud | Azure VM image scans in Azure Government use `management.azure.com` instead of the Azure Government endpoint, which produces `SubscriptionNotFound` errors. Fix pending. | | Jenkins plugin — Missing console output in pipeline job | Pipeline jobs using the Prisma Cloud Compute Jenkins plugin (34.01.119) on Linux Jenkins 2.504.1 succeed and produce scan results in the Console, but the pipeline build console output is empty. Freestyle Docker jobs and Windows Jenkins pipeline jobs are unaffected. | | CSA windows — No metadata details in Manage defenders page | Cloud metadata for Windows Cortex Security Agents (CSA) is shown in Monitor > Hosts but not in Manage > Defenders > Cloud security agents. | | Application control fails to populate several eligible hosts for import | On Defend > Compliance > Hosts > Application control > Import from host, only a small number of eligible hosts appear in the dropdown even when many active defenders are reporting. App-Embedded and Fargate defenders never appear. Under investigation. | | Rhel — Unable to connect rhel container defender to console using podman | RHEL 10 container defenders deployed via Podman log `failed to initialize CRI client: runtime version can not be determined` and do not appear in the Console. | | Automatic deletion of unneeded CVE entries from the intelligence db | Placeholder work item — intelligence DB now automatically removes CVE entries whose Conditions have been null'd by the builder (except RedHat / RedHat-vex and NonVulnerable CVEs). No shipping description yet. | | Move from AWS Go SDK v1 to v2 | Long-running internal migration off the deprecated AWS Go SDK v1 (EOS after 2025-07-31); currently in Review. May not ship in Update 5. | | Fix outdated feeds metric | Internal observability — outdated-feeds metric restoration + threshold tuning. | | Intelligence — Several Amazon CVEs are missing CVSS score in Explore feed | Amazon Linux CVE parsing fallback to older CVSS versions is missing; some Amazon CVEs are reported with score 0 even when NVD has a score. | | Host defender auto-deployment — Add the option to retain the defender installation log | Customer request to retain `twistlock-install.log` after host auto-deploy. Still unassigned. | | Inconsistency with jar packages names | Some JARs are stored under two different package keys (for example, `org.xerial.snappy_snappy-java` vs. `xerial.org_snappy-java`), which causes related CVEs to be missed. Ticket is still `New`; the ticket carries a partial RN description: *"Fixed an issue where old jar versions were not parsed properly."* Confirm shipping status. | | Amazon Linux CVEs — diff between severity/status in Explore vs. ALAS feeds | Intelligence data-quality gap between Amazon's Explore feed (usually authoritative) and the ALAS RSS feed (fallback). No fix — behavior note only. | | CWP API latency degradation — Serverless Functions Evaluate | Latency regression on the Serverless Functions Evaluate route observed between Quinn Update 1 and Quinn Update 2. Related to CRTX-173473. | | Intelligence — Go builder schema issue for MAL-YYYY-NNNN IDs | Go builder rejects OSV IDs of the form `MAL-2025-NNNN` as `Invalid schema ID`. Staging only. | | Intelligence — github: Validation failed for vulnerability.CVEVulnerability | Intelligence staging validation failure for CVE-2023-32198 (rancher/steve). | | Intelligence — Skipping long vector CVSS strings | Intelligence staging logs CVSS 4.0 vector strings >170 chars as errors; should be warnings. | | Intelligence — RockyLinux CVSS missing, fallback to severity | RockyLinux CVSS missing for several CVEs; Intelligence falls back to severity-based calculation. | | Intelligence — pan-os: failed to parse conditions for CVE | pan-os builder fails to parse conditions for CVE-2020-2035, PAN-SA-2025-0005, CVE-2025-0110. | | Intelligence — CVE validation errors: wildcards should be alone in condition | Many Go builder CVEs with wildcard conditions fail validation in Intelligence staging. | | Intelligence — CVE validation error: invalid operator | Intelligence staging rejects `redhat` (oval) CVE-2015-5346 with `version "\"!=1.1-14.el7" does not have a valid operator`. | | Optimization — When downloading from Rocky feeds use more specific url | Rocky Linux errata downloader should filter by `type_security` and by last-updated date. | --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.prismacloud.io/release-notes/look-ahead-planned-updates-on-prisma-cloud/look-ahead-secure-the-runtime.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.