use chroot/umount/pivotroot instead of jailing

[djeebus]

Tasks remaining:

• Resolve race condition (sandbox can't init before adding to map, can't add to map until it's init'd)

Follow up tasks:

• Enable tracing/logging per connection via feature flag
• Use LifecycleID to track sandbox volume mounts rather than SandboxID

---

Note

Medium Risk
Introduces new mount-namespace + pivot_root chroot implementation and changes NFS mount behavior, which can break volume isolation/mount reliability or require elevated privileges in production. Also expands IaC/env configurability, increasing risk of misconfiguration at deploy time.

Overview
Switches the orchestrator’s volume filesystem isolation to a new internal/chrooted implementation that uses a dedicated mount namespace plus mount/pivot_root/umount, and adds a small builder to derive per-team/per-volume paths from PERSISTENT_VOLUME_MOUNTS. The orchestrator config now includes NFS proxy logging/tracing/level env vars (with custom parsing for nfs.LogLevel), envd now waits for all NFS mounts to complete and uses more resilient mount options, and the Terraform/Nomad job wiring is extended to pass arbitrary extra env vars and to make GCP persistent volume types support custom NFS mount options and optional performance (IOPS) config; multiple Go modules also get dependency bumps and mockery generation is moved into the orchestrator package.

^{Written by Cursor Bugbot for commit efdf97b. This will update automatically on new commits. Configure here.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3bce20caf8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[dobrac]

Have you considered using library for at least some parts of the functionality?

(I suppose Kubernetes ecosystem could have pretty good ones)

[dobrac]

Small nits, gtg!

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

There are 3 total unresolved issues (including 2 from previous reviews).

Autofix Details

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Unused nfs_mount_options variable in persistent-volume-types module
  • Removed the unused nfs_mount_options input from the module call and deleted its unreferenced variable declaration from the module.

Or push these changes by commenting:

@cursor push dc2268a65b

Preview (dc2268a65b)

diff --git a/iac/provider-gcp/persistent-volume-types/variables.tf b/iac/provider-gcp/persistent-volume-types/variables.tf
--- a/iac/provider-gcp/persistent-volume-types/variables.tf
+++ b/iac/provider-gcp/persistent-volume-types/variables.tf
@@ -44,8 +44,3 @@
   })
   default = null
 }
-
-variable "nfs_mount_options" {
-  type    = list(string)
-  default = []
-}

diff --git a/iac/provider-gcp/persistent-volumes.tf b/iac/provider-gcp/persistent-volumes.tf
--- a/iac/provider-gcp/persistent-volumes.tf
+++ b/iac/provider-gcp/persistent-volumes.tf
@@ -9,7 +9,6 @@
   location           = each.value.location
   network_name       = var.network_name
   nfs_version        = each.value.nfs_version
-  nfs_mount_options  = each.value.mount_options
   prefix             = var.prefix
   tier               = each.value.tier
   performance_config = each.value.performance_config

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

There are 4 total unresolved issues (including 3 from previous reviews).

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}