Git, Docker & testing

Version control, containers, repeatable tests and deployment diagnostics.

100 troubleshooting guides · Page 3 / 5

49. Why did Compose create another set of containers and volumes?

    Compose uses a project identity to name and group resources. A different directory, project-name option or configuration can create a separate resource set.

50. Does running Docker rootless remove every container security risk?

    Rootless mode reduces some daemon and runtime privileges, but the workload can still access resources explicitly granted to it and can contain vulnerable application code.

51. How can an application run with a read-only container filesystem?

    A read-only root filesystem reveals paths where the application assumes it can write. Temporary files, caches and runtime state need explicit writable locations.

52. Why should internal databases not publish a host port unnecessarily?

    Services on the same container network can communicate without exposing every internal port to the host or external clients. Publishing expands the reachable surface.

53. Why can COPY not access a file outside the Docker build context?

    The build context defines the files available to the build. A Dockerfile path does not automatically grant access to every parent directory on the host.

54. Why does changing runtime environment not alter values baked into a frontend bundle?

    Some build tools replace configuration during compilation. Once bundled, changing a container environment variable does not rewrite those static bytes.

55. How can I check that a deployed container matches the reviewed build?

    A passing source test does not prove the running image was built from that source. Mutable tags and separate rebuilds can break the chain of evidence.

56. Why is disk space not freed after deleting a large Linux log file?

    A process can keep an open file descriptor after the pathname is unlinked. The storage remains in use until the final reference is closed.

57. Why can Linux report no space left when the disk still has free gigabytes?

    A filesystem can exhaust inodes or hit quotas even when data blocks remain free. Many tiny files can consume metadata capacity.

58. Why can I read a file but not access it through its directory path?

    Access requires appropriate permissions on each directory component as well as on the file itself. A parent directory can block traversal.

59. Why are newly created files missing permissions that the application requested?

    The process umask removes permission bits from the requested creation mode. Parent directory policies and ACLs can also affect the result.

60. Why does a program work in my shell but fail as a systemd service?

    A service has a different user, environment, working directory and startup context from an interactive shell. Shell profiles are not a reliable service configuration source.

61. Why does a service stop retrying after several quick failures?

    A service manager can apply start-rate limits to prevent endless rapid restart loops. The limit is a symptom of repeated failure, not necessarily the original cause.

62. Why does a shell pipeline appear successful when an earlier command failed?

    A pipeline’s reported status can come from its final command, allowing an earlier failure to be hidden by a successful formatter or tee process.

63. Why does a shell variable containing spaces turn into several arguments?

    Unquoted expansion can undergo word splitting and pathname expansion. A value intended as one filename can become several arguments or match unrelated files.

64. Why is parsing ls output unsafe for a file-processing script?

    Filenames can contain spaces, tabs, newlines and leading punctuation. Human-oriented listing output is not a reliable record format.

65. Why does a cron job fail although the same command works interactively?

    Cron commonly supplies a smaller environment and a different working directory from your shell. Relative paths, PATH assumptions and missing credentials can surface only there.

66. Why should timeout duration use a monotonic clock rather than wall time?

    Wall time can jump when synchronization or an administrator changes the clock. Measuring elapsed duration with it can produce negative or unexpectedly long intervals.

67. Why does a service report too many open files under load?

    Sockets, files, pipes and other resources consume file descriptors. A leak or an undersized limit can exhaust the process budget.

68. Why can a Linux process remain visible as a zombie after it exits?

    An exited child retains a small process-table record until its parent collects the exit status. The zombie is not continuing to execute application code.

69. How do I diagnose address already in use without killing random processes?

    Another listener may own the requested address and port, or a duplicate application instance may already be running. The conflicting address family also matters.

70. Why can rolling back application code fail after a database migration?

    The old binary may no longer understand a schema changed by the new release. A code rollback and a data rollback are different operations.

71. Why does a test pass alone but fail when the whole suite runs?

    Another test may leave shared state behind, or the test may depend on execution order, environment, time or resource ownership. Isolation failures often hide in full-suite runs.

72. How can a session-scoped test fixture accidentally share mutable state?

    A broad fixture scope reuses the same object across many tests. If one test mutates it, later tests may start with a different baseline.

Ask your own question · Sign in to post questions and answers.