Frequently Asked Questions
This appendix provides contact information and answers to frequently asked questions.
How do I post an issue related to Open CAS Linux
Post a new issue at the following URL:
- Open CAS GitHub Issue (https://github.com/Open-CAS/open-cas-linux/issues/new)
Why does Open CAS Linux use some DRAM space?
Open CAS Linux uses a portion of system memory for metadata, which tells us where data resides. The amount of memory needed is proportional to the size of the cache space. This is true for any caching software solution. However with Open CAS Linux this memory footprint can be decreased using a larger cache line size set by the parameter --cache-line-size which may be useful in high density servers with many large HDDs.
How do I test performance?
In addition to the statistics provided (see Monitoring Open CAS Linux for details), third-party tools are available that can help you test IO performance on your applications and system, including:
dt (https://github.com/RobinTMiller/dt) for disk access simulations
Is it possible to experience slower than HDD performance when using caching?
Yes, it is possible. For example, if the cache is in write-back mode and the entire cache is full of dirty data and a read occurs which requires new blocks to be loaded into the cache, performance will be degraded even if the read is sequential. The cache must first evict dirty blocks, which requires random writes to the HDD, then read the new data from the HDD and finally, write it to the cache. Whereas, without caching it would have simply resulted in a single read from the HDD. To avoid situations such as these, Open CAS Linux opportunistically flushes dirty data from the cache during idle IO times.
Where are the cached files located?
Open CAS Linux does not store files on disk; it uses a pattern of blocks on the SSD as its cache. As such, there is no way to look at the files it has cached.
How do I delete all the Open CAS Linux installation files?
Stop the Open CAS Linux software as described in the section Stopping Cache Instances.
If Open CAS Linux was installed via the
make install command, Open CAS Linux can be uninstalled using the command:
# make uninstall
Otherwise, if Open CAS Linux was manually installed then manually unload any CAS kernel modules. Once uninstalled remove the source files.
Does Open CAS Linux support write-back caching?
Yes. See the section Manual Configuration for Write-Back Mode for details.
Must I stop caching before adding a new pair of cache/core devices?
No, you can create new cache instances while other instances are running.
Can I assign more than one core device to a single cache?
Yes. Many core devices (up to 4096) may be associated with a single cache drive or instance. You can
add them using the
casadm -A command.
Can I add more than one cache to a single core device?
No, if you want to map multiple cache devices to a single core device, the cache devices must appear as a single block device through the use of a system such as RAID-0.
Why do tools occasionally report data corruption with Open CAS Linux?
Some applications, especially micro benchmarks like dt and FIO, may use a device to perform direct or raw accesses. Some of these applications may also allow you to configure values like a device’s alignment and block size restrictions explicitly, for instance via user parameters (rather than simply requesting these values from the device). In order for these programs to work, the block size and alignment for the cache device must match the block size and alignment selected in the tool.
Do I need to partition the cache device?
No. If you do not specify a partition, Open CAS Linux uses the entire device as the cache device.
Can I use a partition on a SSD as a cache device?
Yes, however, using the entire SSD device as the cache is highly recommended for best performance.
Do I need to format the partition or the device configured as the cache device?
No, the cache device has no format requirement. However, note that Open CAS Linux overwrites any data previously stored on the caching device (including any partition formatting).
What is the logical and physical block size for Open CAS Linux cache volumes (for exported objects)?
The logical block size for Open CAS Linux cache volumes is inherited from the core device, while the physical block size will be represented as the larger of the physical block sizes of the cache or core devices.
It is not possible to add a core device to a cache instance when the logical block size of the cache device is greater than that of the core device (eg. when the SSD has 4KiB logical block size and the HDD has 512B logical block size).
What happens if my SSD or HDD becomes unresponsive or disconnected?
In the event of a cache or core device becoming unresponsive, Open CAS Linux will fail all IO to all exported devices for the related cache (eg. /dev/cas1-1, /dev/cas1-2, etc.). To resume IO to the exported devices for the given cache, the user must restart the affected cache (in this example, cache ID 1).
When my device becomes disconnected, will I receive any notification from Cache Acceleration Software (CAS)?
No. The OS does not send notification to Open CAS Linux of the disconnection of the device, so the software will not know of this event until IO to the device is attempted. The device will still be listed in the --list-caches and --stats output, and no warning will be logged, until IO to the device is attempted. Check /var/log/messages and dmesg for standard Linux device IO errors.
Where is the log file located?
All events are logged in the standard Open CAS Linux system logs. Use the dmesg command or inspect the /var/log/messages file.
To log all messages during testing or kernel debugging, use the command:
# echo 8 > /proc/sys/kernel/printk.
Typical log sample of successful cache initialization:
Cache line size: 64 KiB
Metadata capacity: 25 MiB
Parameters (policies) accepted:: 0 1 1 4
Pre-existing metadata, Clean shutdown
Done saving cache state!
Cache 1 successfully added
IO Scheduler of cas1-1 is cfq
Core “/dev/disk/by-id/wwn-0x50014ee004276c68” successfully added
Typical log sample of successful cache removal:
Removing Cache 1
Trying to remove 1 cached device(s)
Removed device cas1-1.
Done saving cache state!
Cache 1 successfully removed
Why does flushing the cache drive in Write Back mode take so long?
Flushing time has many factors, including but not limited to, cache device capacity, storage performance, and overall system utilization. Flushing time can vary greatly depending on these factors. You can check the status of the cache device being flushed by using the -L option in casadm.
# casadm -L
The following command can be used as well to verify activity and loading of the IO devices.
# iostat -xmt 1
Why does the CAS statistic counter for “Reads from Cache” increase when only write operations are being issued from the host application?
The Reads from Cache counter from the CAS statistic output ( See Monitoring CAS ) can increase when only doing writes from the host in several cases. For example, during the flushing of dirty data to the backing store in write-back or write-only mode. Similarly, during eviction when the cache is full in write-back or write-only mode and there is in incoming write which requires to read dirty data from the cache and flush it to backing store in order to make space for the incoming write. Another example is if there are any misaligned I/O that requires read-modify-write of a block that is in the cache.
Should nr_request be modified for further tuning?
The nr_requests controls how many requests may be allocated in the block layer for read or write requests. It represents the IO queue size. Each request queue has a limit on the total number of request descriptors that can be allocated for each read and write IO. By default, the number is 128, meaning 128 reads and 128 writes can be queued at a time before putting a process to sleep.
Large values of nr_request may increase throughput for workloads writing many small files. For larger IO operations, you may decrease the nr_request value. To get better read performance, you can set the nr_request value to 1024 for example, but increasing the value too high might introduce latency and degrade write performance. For latency sensitive applications, the converse is also true.
Configuring nr_requests to a larger value may improve performance for some customers who are caching multi-HDD RAID onto one SSD. However, this may not be useful with environments, such as Ceph, where each HDD is being cached to one SSD or a partition on an SSD. Your application and its IO patterns may vary and require a detailed tuning exercise.
To change the nr_requests value, use the following procedure:
# cat /sys/block/sda/queue/nr_requests
# echo 256 > /sys/block/sda/queue/nr_requests