github-actions[bot] commented on issue #14326: URL: https://github.com/apache/cloudstack/issues/14326#issuecomment-6039198112
## ๐ฏ Triage report Confirmed bug: `kvmsmpheartbeat.sh` (added by #12773) requires the SharedMountPoint pool path itself to be a mount point, but subdirectories of clustered/parallel filesystems (GPFS, OCFS2, GFS2, CephFS) are a valid and documented configuration. This breaks VM-HA heartbeat writes and, combined with `reboot.host.and.alert.management.on.heartbeat.timeout=true`, can trigger simultaneous hard resets across a whole cluster. A maintainer (`@weizhouapache`) already confirmed it as a valid bug, and the author has tested a fix and opened PR #14331. ### ๐ Assessment | Dimension | Value | Reasoning | |---|---|---| | **Type** | type:bug | Regression in HA heartbeat logic introduced by #12773, confirmed by a maintainer. | | **Component** | component:kvm, component:primary-storage | Affects KVM agent HA heartbeat script for SharedMountPoint primary storage. | | **Severity** | Severity:Major | Can cause cluster-wide, simultaneous host hard-resets and storage quorum loss in hyperconverged setups; workaround exists (bind-mount). | | **Labels** | type:bug, component:kvm, component:primary-storage, Severity:Major | See above | | **Coding agent** | Not suitable | Fix already authored, tested, and submitted as PR #14331 by the reporter; no further automated work needed here. | ### ๐ Similar issues - https://github.com/apache/cloudstack/issues/14328 (duplicate, already closed as not_planned) โ identical report by the same author, closed in favor of this one. - https://github.com/apache/cloudstack/issues/14327 (related) โ broader enhancement request for first-class parallel/clustered filesystem support, references this bug as a concrete example. - https://github.com/apache/cloudstack/issues/12773 (related) โ PR that introduced the SharedMountPoint heartbeat feature and the regression. <details><summary>๐ก Notes and suggestions</summary> - PR #14331 reportedly also fixes a related issue in `check_hbLog`: `expr` exits with status 1 when the computed diff is 0, causing a live host to be misreported as DEAD if read happens in the same second as a write. - The author's testing (per the comments) covered: subdirectory-of-GPFS-mount heartbeats, real mount points (tmpfs), NFS pools, refusal on root-FS paths and nonexistent paths, and a live 3-host HA failover test with sysrq hard reset โ all passing with the patch. - One side issue noted during failover testing: a killed heartbeat write can leave behind an empty `hb-<ip>.<pid>` temp file; the author offered to add a `trap` cleanup to the PR โ maintainers may want to confirm this is included before merge. - Recommend maintainers prioritize review of PR #14331 given the severity (cluster-wide reboot risk) and that it's already tested. </details> > Generated by [Daily Issue Triage](https://github.com/apache/cloudstack/actions/runs/37629898565) ยท sonnet50 92.1K ยท [โท](https://github.com/search?q=repo%3Aapache%2Fcloudstack+%22gh-aw-workflow-call-id%3A+apache%2Fcloudstack%2Fdaily-issue-triage%22&type=issues) > <details> <summary>Add this agentic workflows to your repo</summary> To install this agentic workflow, run ``` gh aw add githubnext/agentics/workflows/daily-issue-triage.md@d7c1dc4b72b00607a67caaffdcc216cb64379cf9 ``` </details> <!-- gh-aw-agentic-workflow: Daily Issue Triage, engine: copilot, version: 1.0.52, model: claude-sonnet-5, id: 37629898565, workflow_id: daily-issue-triage, run: https://github.com/apache/cloudstack/actions/runs/37629898565 --> <!-- gh-aw-workflow-call-id: apache/cloudstack/daily-issue-triage --> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
