| 1 | <?xml version='1.0' encoding='UTF-8'?>
|
|---|
| 2 | <!DOCTYPE topic PUBLIC "-//OASIS//DTD DITA Topic//EN" "topic.dtd">
|
|---|
| 3 | <topic xml:lang="en-us" id="ts_config-periodic-flush">
|
|---|
| 4 | <title><?oxy_comment_start author="Mhari Duncan" timestamp="20241210T152506+0000" comment="kespenla: This is not a true "known issue", it is documenting that sluggish I/O on the host can cause I/O timeouts in the VM which may have usability problems. This is quite old (but stll true) information. Likely not as frequently seen as when this was originally written 15+ years ago."?>Guest Shows IDE/SATA Errors for File-Based Images on Slow Host File System<?oxy_comment_end?></title>
|
|---|
| 5 |
|
|---|
| 6 | <body>
|
|---|
| 7 | <p>
|
|---|
| 8 | Occasionally, some host file systems provide very poor writing
|
|---|
| 9 | performance and as a consequence cause the guest to time out
|
|---|
| 10 | IDE/SATA commands. This is normal behavior and should normally
|
|---|
| 11 | cause no real problems, as the guest should repeat commands that
|
|---|
| 12 | have timed out. However, guests such as some Linux versions have
|
|---|
| 13 | severe problems if a write to an image file takes longer than
|
|---|
| 14 | about 15 seconds. Some file systems however require more than a
|
|---|
| 15 | minute to complete a single write, if the host cache contains a
|
|---|
| 16 | large amount of data that needs to be written.
|
|---|
| 17 | </p>
|
|---|
| 18 | <p>
|
|---|
| 19 | The symptom for this problem is that the guest can no longer
|
|---|
| 20 | access its files during large write or copying operations,
|
|---|
| 21 | usually leading to an immediate hang of the guest.
|
|---|
| 22 | </p>
|
|---|
| 23 | <p>
|
|---|
| 24 | In order to work around this problem, the true fix is to use a
|
|---|
| 25 | faster file system that does not exhibit such unacceptable write
|
|---|
| 26 | performance, it is possible to flush the image file after a
|
|---|
| 27 | certain amount of data has been written. This interval is
|
|---|
| 28 | normally infinite, but can be configured individually for each
|
|---|
| 29 | disk of a VM.
|
|---|
| 30 | </p>
|
|---|
| 31 | <p>
|
|---|
| 32 | For IDE disks use the following command:
|
|---|
| 33 | </p>
|
|---|
| 34 | <pre xml:space="preserve">VBoxManage setextradata <varname>VM-name</varname>
|
|---|
| 35 | "VBoxInternal/Devices/piix3ide/0/LUN#[<varname>x</varname>]/Config/FlushInterval" [<varname>b</varname>]</pre>
|
|---|
| 36 | <p>
|
|---|
| 37 | For SATA disks use the following command:
|
|---|
| 38 | </p>
|
|---|
| 39 | <pre xml:space="preserve">VBoxManage setextradata <varname>VM-name</varname>
|
|---|
| 40 | "VBoxInternal/Devices/ahci/0/LUN#[<varname>x</varname>]/Config/FlushInterval" [<varname>b</varname>]</pre>
|
|---|
| 41 | <p><codeph>[<varname>x</varname>]</codeph> specifies the
|
|---|
| 42 | disk. For IDE, <codeph>0</codeph> represents device 0 on the
|
|---|
| 43 | primary channel, <codeph>1</codeph> represents device 1 on the
|
|---|
| 44 | primary channel, <codeph>2</codeph> represents device 0 on the
|
|---|
| 45 | secondary channel, and <codeph>3</codeph> represents device 1
|
|---|
| 46 | on the secondary channel. For SATA, use values between
|
|---|
| 47 | <codeph>0</codeph> and <codeph>29</codeph>. This
|
|---|
| 48 | configuration option applies to disks only. Do not use this
|
|---|
| 49 | option for CD or DVD drives.
|
|---|
| 50 | </p>
|
|---|
| 51 | <p>
|
|---|
| 52 | The unit of the interval
|
|---|
| 53 | (<codeph>[<varname>b</varname>]</codeph>) is the
|
|---|
| 54 | number of bytes written since the last flush. The value for it
|
|---|
| 55 | must be selected so that the occasional long write delays do not
|
|---|
| 56 | occur. Since the proper flush interval depends on the
|
|---|
| 57 | performance of the host and the host filesystem, finding the
|
|---|
| 58 | optimal value that makes the problem disappear requires some
|
|---|
| 59 | experimentation. Values between 1000000 and 10000000 (1 to 10
|
|---|
| 60 | megabytes) are a good starting point. Decreasing the interval
|
|---|
| 61 | both decreases the probability of the problem and the write
|
|---|
| 62 | performance of the guest. Setting the value unnecessarily low
|
|---|
| 63 | will cost performance without providing any benefits. An
|
|---|
| 64 | interval of 1 will cause a flush for each write operation and
|
|---|
| 65 | should solve the problem in any case, but has a severe write
|
|---|
| 66 | performance penalty.
|
|---|
| 67 | </p>
|
|---|
| 68 | <p>
|
|---|
| 69 | Providing a value of <codeph>0</codeph> for
|
|---|
| 70 | <codeph>[<varname>b</varname>]</codeph> is treated as
|
|---|
| 71 | an infinite flush interval, effectively disabling this
|
|---|
| 72 | workaround. Removing the extra data key by specifying no value
|
|---|
| 73 | for <codeph>[<varname>b</varname>]</codeph> has the
|
|---|
| 74 | same effect.
|
|---|
| 75 | </p>
|
|---|
| 76 | </body>
|
|---|
| 77 |
|
|---|
| 78 | </topic>
|
|---|
