[Kimchi-devel] [PATCH v2] host/partitions: avoid calling disks.get_partitions_names() for each partition
Zhou Zheng Sheng
zhshzhou at linux.vnet.ibm.com
Fri May 9 01:54:07 UTC 2014
on 2014.05/08 23:08, Aline Manera wrote:
> On 05/07/2014 03:17 AM, Zhou Zheng Sheng wrote:
>> Problem:
>> In PartitionModel.lookup(), it calls disks.get_partitions_names() to
>> verify the given partition is eligible to be used in a new logical
>> storage pool as follow.
>> if name not in disks.get_partitions_names():
>> raise NotFoundError(...)
>>
>> When the front-end issues GET request to host/partitions, the back-end
>> would call get_partitions_names() for each partition. If there are many
>> disks and partitions in the host, get_partitions_names() takes a long
>> time to finish, and calling get_partitions_names() for each partition
>> would make the GET request timeout.
>>
>> In all, the original control flow is as follow.
>> GET host/partitions
>> Partitions._get_resources()
>> PartitionsModel.get_list()
>> disks.get_partitions_names()
>> for each partition name:
>> Partition.lookup()
>> PartitionModel.lookup()
>> verify partition name in disks.get_partitions_names()
>> disks.get_partition_details()
>>
>> The solution is to split the block device verification code out of
>> disks.get_partitions_names(), and put it in a new function
>> disks._is_available(). Now get_partitions_names() just returns
>> all the block devices, and disks.get_partition_details() can perform
>> the eligibility check by calling disks._is_available(). The original
>> get_partitions_names() checks eligibility for all the detected device,
>> so calling it for each partition is too slow. The new _is_available()
>> function just checks the given block device.
>>
>> The New control flow is as follow.
>> GET host/partitions
>> Partitions._get_resources()
>> PartitionsModel.get_list()
>> disks.get_partitions_names()
>> for each partition name:
>> Partition.lookup()
>> PartitionModel.lookup()
>> disks.get_partition_details()
>> check if the block device _is_available()
>>
>> v1:
>> Make use of the resource and model arguments to have Partitions
>> collection controller set a flag, so Partition model can inspect the
>> flag and skip the check accordingly.
>>
>> v2:
>> v1 solution is too hack, re-write the patch by refactoring the
>> underlying disks model. Split the batch verification into small peices
>> amortized by each concrete block device.
>>
>> Signed-off-by: Zhou Zheng Sheng <zhshzhou at linux.vnet.ibm.com>
>> ---
>> src/kimchi/control/host.py | 6 +++++-
>> src/kimchi/disks.py | 37 +++++++++++++++++++++++--------------
>> src/kimchi/model/host.py | 3 ---
>> 3 files changed, 28 insertions(+), 18 deletions(-)
>>
>> diff --git a/src/kimchi/control/host.py b/src/kimchi/control/host.py
>> index ad34919..ebf1bed 100644
>> --- a/src/kimchi/control/host.py
>> +++ b/src/kimchi/control/host.py
>> @@ -21,7 +21,7 @@ import cherrypy
>>
>> from kimchi.control.base import Collection, Resource, SimpleCollection
>> from kimchi.control.utils import UrlSubNode, validate_method
>> -from kimchi.exception import OperationFailed
>> +from kimchi.exception import OperationFailed, NotFoundError
>> from kimchi.template import render
>>
>>
>> @@ -80,6 +80,7 @@ class Partitions(Collection):
>> # sorted by their path
>> def _get_resources(self, flag_filter):
>> res_list = super(Partitions, self)._get_resources(flag_filter)
>> + res_list = filter(lambda x: x.info['available'], res_list)
>> res_list.sort(key=lambda x: x.info['path'])
>> return res_list
>>
>> @@ -90,6 +91,9 @@ class Partition(Resource):
>>
>> @property
>> def data(self):
>
>> + if not self.info['available']:
>> + raise NotFoundError("KCHPART0001E", {'name':
>> self.info['name']})
>> +
>
> This logic should be in model
>
Thanks Aline, you are correct. Originally this is in model, but it
brings a problem.
This patch is to avoid redundant disks._is_available() invocation.
Before this patch, the availability checking is in both when Kimchi
generates the partition names and when Kimchi lookups the particular
partition info. So When the front-end GETs host/partitions, Kimchi
generates a list of partition names, then lookups the information, and
each partition gets checked twice, thus makes the process slow.
So my idea is to drop the availability check in the
disks.get_partitions_names(), so it returns all partitions, then we
check when it lookups the partition info. However, if it raises
NotFoundError in PartitionModel.lookup(), Partitions._get_resources()
would explode because the partition names returned from
disks.get_partitions_names() may contain unavailable partitions.
So I have two choices to deal with this problem. I can either drop the
invocation to super._get_resources() and complete re-implement
Partitions._get_resources() to catch the NotFoundError exception raised
by PartitionModel.lookup(), or I have it do not raise the exception but
set an availability flag for later check. I think the second choice is
less evil.
>> return self.info
>>
>>
>> diff --git a/src/kimchi/disks.py b/src/kimchi/disks.py
>> index 842fd48..a13c753 100644
>> --- a/src/kimchi/disks.py
>> +++ b/src/kimchi/disks.py
>> @@ -142,7 +142,23 @@ def _get_vgname(devNodePath):
>> return re.findall(r"LVM2_VG_NAME='([^\']*)'", out)[0]
>>
>>
>> -def get_partitions_names():
>> +def _is_available(name, devtype, fstype, mountpoint, majmin):
>> + devNodePath = _get_dev_node_path(majmin)
>> + # Only list unmounted and unformated and leaf and (partition or
>> disk)
>> + # leaf means a partition, a disk has no partition, or a disk not
>> held
>> + # by any multipath device. Physical volume belongs to no volume
>> group
>> + # is also listed. Extended partitions should not be listed.
>> + if (devtype in ['part', 'disk', 'mpath'] and
>> + fstype in ['', 'LVM2_member'] and
>> + mountpoint == "" and
>> + _get_vgname(devNodePath) == "" and
>> + _is_dev_leaf(devNodePath) and
>> + not _is_dev_extended_partition(devtype, devNodePath)):
>> + return True
>> + return False
>> +
>> +
>> +def get_partitions_names(check=False):
>> names = set()
>> keys = ["NAME", "TYPE", "FSTYPE", "MOUNTPOINT", "MAJ:MIN"]
>> # output is on format key="value",
>> @@ -151,26 +167,17 @@ def get_partitions_names():
>> # split()[0] to avoid the second part of the name, after the
>> # whiteline
>> name = dev['name'].split()[0]
>> - devNodePath = _get_dev_node_path(dev['maj:min'])
>> - # Only list unmounted and unformated and leaf and (partition
>> or disk)
>> - # leaf means a partition, a disk has no partition, or a disk
>> not held
>> - # by any multipath device. Physical volume belongs to no
>> volume group
>> - # is also listed. Extended partitions should not be listed.
>> - if not (dev['type'] in ['part', 'disk', 'mpath'] and
>> - dev['fstype'] in ['', 'LVM2_member'] and
>> - dev['mountpoint'] == "" and
>> - _get_vgname(devNodePath) == "" and
>> - _is_dev_leaf(devNodePath) and
>> - not _is_dev_extended_partition(dev['type'],
>> devNodePath)):
>> + if check and not _is_available(name, dev['type'], dev['fstype'],
>> + dev['mountpoint'],
>> dev['maj:min']):
>> continue
>> -
>> names.add(name)
>>
>> return list(names)
>>
>>
>> def get_partition_details(name):
>> - dev_path = _get_dev_node_path(_get_dev_major_min(name))
>> + majmin = _get_dev_major_min(name)
>> + dev_path = _get_dev_node_path(majmin)
>>
>> keys = ["TYPE", "FSTYPE", "SIZE", "MOUNTPOINT"]
>> try:
>> @@ -180,6 +187,8 @@ def get_partition_details(name):
>> "Error getting partition info for %s: %s", name, e)
>> return {}
>>
>> + dev['available'] = _is_available(name, dev['type'], dev['fstype'],
>> + dev['mountpoint'], majmin)
>> if dev['mountpoint']:
>> # Sometimes the mountpoint comes with [SWAP] or other
>> # info which is not an actual mount point. Filtering it
>> diff --git a/src/kimchi/model/host.py b/src/kimchi/model/host.py
>> index 47b0aa1..e9ac487 100644
>> --- a/src/kimchi/model/host.py
>> +++ b/src/kimchi/model/host.py
>> @@ -245,9 +245,6 @@ class PartitionModel(object):
>> pass
>>
>> def lookup(self, name):
>> - if name not in disks.get_partitions_names():
>> - raise NotFoundError("KCHPART0001E", {'name': name})
>> -
>> return disks.get_partition_details(name)
>>
>>
>
--
Zhou Zheng Sheng / 周征晟
E-mail: zhshzhou at linux.vnet.ibm.com
Telephone: 86-10-82454397
More information about the Kimchi-devel
mailing list