**Container** başlatılıp çalıştığında (bir **container image**'dan başlatılır), dosyalar oluşturabilir/değiştirebilir, environment variable'ları değiştirebilir vb. Bu değişiklikler sadece o container içinde kalır; alttaki container image'da kalıcı olmaz (diske kaydedilmez).

Bir container image, **program** dosyası ve içeriklerine benzetilebilir; örn. `python` ve `main.py` gibi bir dosya.

   *     READ 335c4c6028171cfddfbaae1a9c313c52
   *     READ 3400330d1dfc7f3f7f4b8d4d803dfcf6
   *
   * The first five lines of the journal form its header. They are the constant string
   * "libcore.io.DiskLruCache", the disk cache's version, the application's version, the value
   * count, and a blank line.
   *
   * Each of the subsequent lines in the file is a record of the state of a cache entry. Each line

| `minio_node_drive_writes_await`      | Average time for write requests to be served on a drive. |
| `minio_node_drive_perc_util`         | Percentage of time the disk was busy since uptime.       |

## Network Interface Metrics

| Name                          | Description                                                |

 *  Fix: Don't crash decompressing web sockets messages. We had a bug where we assumed deflated
    bytes in would always yield deflated bytes out and this isn't always the case!

 *  Fix: Reliably update and invalidate the disk cache on windows. As originally designed our
    internal `DiskLruCache` assumes an inode-like file system, where it's fine to delete files that

		for _, v := range vectors {
			b.Run(v.label, func(b *testing.B) {
				b.ReportAllocs()
				for i := 0; i < b.N; i++ {
					// Writing to io.Discard because we want to
					// test purely the writer code and not bring in disk performance into this.
					tw := NewWriter(io.Discard)
					for _, file := range v.files {
						if err := tw.WriteHeader(file.hdr); err != nil {
							b.Errorf("unexpected WriteHeader error: %v", err)
						}

   *       implementation will effectively put the whole file in memory. On many systems with paging
   *       and virtual memory, this is not a problem - because it is mapped read-only, the kernel
   *       can always page it to disk "for free". However, on systems where killing processes
   *       happens all the time in normal conditions (i.e., android) the OS must make a tradeoff

configuration `criSocket` field or the `--cri-socket` CLI flag. Automatically add a missing URL scheme to the user configuration in memory, but warn them that they should also update their configuration on disk manually. During `kubeadm upgrade apply/node` mutate the `/var/lib/kubelet/kubeadm-flags.env` file on disk and the `kubeadm.alpha.kubernetes.io/cri-socket` annotation Node object if needed. These automatic actions are temporary and will be removed in a future release. In the future the kubelet may not...

A container image is comparable to the **program** file and contents, e.g. `python` and some file `main.py`.

func isMaxPartID(partID int) bool {
	return partID > globalMaxPartID
}

// profilerWrapper is created because pkg/profiler doesn't
// provide any API to calculate the profiler file path in the
// disk since the name of this latter is randomly generated.
type profilerWrapper struct {
	// Profile recorded at start of benchmark.
	records map[string][]byte
	stopFn  func() ([]byte, error)
	ext     string
}

		// non-zero and only the second line above, the TOC,
		// is what matters.
	})
}

// rZipBytes returns the bytes of a recursive zip file, without
// putting it on disk and triggering certain virus scanners.
func rZipBytes() []byte {
	s := `
0000000 50 4b 03 04 14 00 00 00 08 00 08 03 64 3c f9 f4
0000010 89 64 48 01 00 00 b8 01 00 00 07 00 00 00 72 2f