Skip to main content

40 posts tagged with "info"

View All Tags

· One min read
Joachim Kraftmayer

When commissioning new Ceph clusters, our standard tests also include measuring the I/O latency for RBD.

We also always measure the performance values for the entire stack. Over the years, we have seen the results of our hard work in improving ceph osd in various tests.

For our tests, we create a temporary work file and read random blocks with non-cached read operations from it.

We are now measuring latencies of 300 to 600 microseconds.

· One min read
Joachim Kraftmayer

In the presentation by Sage Weil at the Openstack Summit 2018 in Berlin, he presents the roadmap for the upcoming features in version Nautilus and the following. Strong focus on the trend of multi- and hybrid cloud environments and scaling of Ceph clusters across data center boundaries.

www.slideshare.net/sageweil1/ceph-data-services-in-a-multi-and-hybrid-cloud-world

· 2 min read
Joachim Kraftmayer

Again and again I come across people who use a replication of 2 for a replicated Ceph pool.

If you know exactly what you are doing, you can do this. I would strongly advise against it.

One simple reason for this is that you can't form a clear majority with two people, there always has to be at least a third.

There are error scenarios in which it can quickly happen that both OSDs (osdA and osdB) of a placement group (replication size = 2) are not available. If an osdA fails, the cluster only has one copy of the object and the default value (min_size = 2) on the pool means that the cluster would no longer allow any write operations to the object.

If min_size=1 (not recommended) then the osdB could be gone for a short time and the osdA would come back. Now osdA does not know whether further write operations have been carried out on osdB during its offline phase.

You can then only hope that all osds come back or you can then manually make the decision for the most current data set. While in the background more and more blocked_requests accumulate in the cluster that would like to access the data.

· 3 min read
Joachim Kraftmayer

With the default options you will see blocked requests in the cluster caused by deep-scrubing operations

recomended deep scrub options to minimize the impact of scrub/deep-scrub in the ceph cluster​

the following options define the scrub and deep-scrub behaviour, by cpu load, osd scheduler priority, defined check intervals and ceph cluster health state

[osd]
#reduce scrub impact
osd max scrubs = 1
osd scrub during recovery = false
osd scrub max interval = 4838400 # 56 days
osd scrub min interval = 2419200 # 28 days
osd deep scrub interval = 2419200
osd scrub interval randomize ratio = 1.0
# osd deep scrub randomize ratio = 1.0
osd scrub priority = 1
osd scrub chunk max = 1
osd scrub chunk min = 1
osd deep scrub stride = 1048576 # 1 MB
osd scrub load threshold = 5.0
osd scrub sleep = 0.3
osd max scrubs
osd max scrubs

Description: The maximum number of simultaneous scrub operations for a Ceph OSD Daemon.

Type: 32-bit Int

Default: 1
osd scrub during recovery

Description: Allow scrub during recovery. Setting this to false will disable scheduling new scrub (and deep–scrub) while there is active recovery. Already running scrubs will be continued. This might be useful to reduce load on busy clusters.

Type: Boolean

Default: true
osd scrub min interval

Description: The minimal interval in seconds for scrubbing the Ceph OSD Daemon when the Ceph Storage Cluster load is low.

Type: Float

Default: Once per day. 60*60*24

5,14 6%
osd scrub interval randomize ratio

Description: Add a random delay to osd scrub min interval when scheduling the next scrub job for a placement group. The delay is a random value less than osd scrub min interval * osd scrub interval randomized ratio. So the default setting practically randomly spreads the scrubs out in the allowed time window of [1, 1.5] * osd scrub min interval.

Type: Float

Default: 0.5
osd scrub priority

Description: The priority set for scrub operations. It is relative to osd client op priority.

Type: 32-bit Integer

Default: 5

Valid Range: 1-63
osd scrub chunk max

Description: The maximum number of object store chunks to scrub during single operation.

Type: 32-bit Integer

Default: 25
osd scrub chunk min

Description: The minimal number of object store chunks to scrub during single operation. Ceph blocks writes to single chunk during scrub.

[docs.ceph.com/docs/master/rados/configuration/osd-config-ref/](http://docs.ceph.com/docs/master/rados/configuration/osd-config-ref/)

Default: 5
osd deep scrub stride

Description: Read size when doing a deep scrub.

Type: 32-bit Integer

Default: 512 KB. 524288
osd scrub load threshold

Description: The normalized maximum load. Ceph will not scrub when the system load (as defined by getloadavg() / number of online cpus) is higher than this number. Default is 0.5.

Type: Float

Default: 0.5
osd scrub sleep

Description: Time to sleep before scrubbing next group of chunks. Increasing this value will slow down whole scrub operation while client operations will be less impacted.

Type: Float

Default: 0

SOURCES​

http://docs.ceph.com/docs/master/rados/configuration/osd-config-ref/

https://indico.cern.ch/event/588794/contributions/2374222/attachments/1383112/2103509/Configuring_Ceph.pdf

https://github.com/ceph/ceph/blob/master/src/common/options.cc#L3130