CCU - Cache Control Utility
CCU is designed to load data into the cache, manage its content and collect statistics. The utility supports a flexible configuration which includes all the specified parameters.
The main configuration file is the /etc/ccu/ccu.conf. The configuration in this file is specified using the YAML format.
The basic elements of the YAML format are:
default: offpeak- "00:00:00 - 01:00:00"- "17:00:00 - 23:59:59"It is possible to create complex data structures based on the basic elements.
One of the principles describing complex structures laid down in the YAML format is the formatting nested elements rule using indents from the beginning of the line. Elements located on the same hierarchy level must have the same number of leading spaces, for example, the following text:
# # Time classes time_classes: # Defines the classes of time default: offpeak # Specifies the default value for time classes # All others items define the names of time classes and rules for them definition # Each rule is started from the name of day category and contains # the list of ranges of times in form "HH:MI:SS-HH:MI:SS" peak: workdays: - "00:00:00 - 01:00:00" - "17:00:00 - 23:59:59" weekend_eve: - "00:00:00 - 02:00:00" - "16:00:00 - 23:59:59" weekend: - "00:00:00 - 02:00:00" - "09:00:00 - 23:59:59" holidays: - "00:00:00 - 02:00:00" - "09:00:00 - 23:59:59"
can be treated as the time_classes data structure having the following fields: default and peak. The peak field in turn is also a structure with the workdays, weekend_eve, weekendand holidays fields which are the named string series.
Given the above, the main rule when editing YAML format documents is using a certain number of whitespaces to form indents instead of using tabs.
time_classes/peak/workdaystime_classes/peak/<the_day_category>time_classes/peak/workdays, or when describing the full name, you can specify its base, for example: the time classes description (time_classes) are specified by the parameters in the <time_class>/<the_day_category>form that contain a listing of time intervals
The following two parameters specify the directories where files with running processes identifiers will be stored -pid_files_path and working files of the cache management utility - work_files_path. Among the working files, in particular, are the files that store the cache usage statistics.
# # PID files path, default: /var/run/ccu pid_files_path: /var/run/ccu # # Work files directory, default: /var/lib/ccu work_files_path: /var/lib/ccu
When the certain events occur the ccu can perform actions specified by the user. Actions to be executed should be specified to meet the rules of the command line interpreter.
Only one event currently is supported: creation of a text file containing information about objects stored in caches. To set the response to the event, you should define the events/on_after_enumeration_creation parameter. A command to create a binary cache contents representation in the event configuration template is specified.
# # Events events: # Command which should be executed after caches' files enumeration on_after_enumeration_creation: "cut -d' ' -f2 /var/cache/ccu/data/enumerated.cs | url2dic /var/cache/ccu/data/enumerated.bin"
Logging options include the path to the log files, as well as the message levels to be written to the file. The most detailed journaling level is the debug level, i.e. when it is enabled all possible messages will be written to the file. When the error journaling level is specified, only errors will be written to the file.
The logging levels are specified separately for each ccu command.
# # Logging parameters logging: path: /var/log/ccu levels: # Enabled log levels for particular command # Possible levels are: # error, warning, info, diagnostic, debug # default logging level is "info" load: "info" purge: "info" remove: "warning" online: "info" monitor: "info"
The fields of the statistics/collectors/
parameter specify the node address and the port the data about the cache usage will be sent. In the future, the name of the statistical collector can be used when describing the CACHEs.
# # Statistics parameters statistics: collectors: # list of collectors' descriptions local: # name of collector description host: localhost # listening address port: 1600 # listening port
Day classes and time classes are used to specify the restrictions for data loading to the cache.
In order to specify the category of the day, you should create the day_categories/<day_category> parameter, whose value is the list of days that fall into the created day category.
The day list element can be both the name of the day of the week, as well as a partially or fully defined date.
When determining the category of a day for a specific date, the more particular definitions are examined in the first place, after that the more general ones are examined.
# # Day categories day_categories: # Defines day categories which could be used in time class definitions # Items under "day_categories" define names of categories; # To define the day categorie it is necessary to specify list of values, # value can be one of following: # - week day # - partially or fully defined date in form DD.MM.YYYY # for example: # "01" - the first day of each month # "07.01" - the 7th of January of each year # "09.05.2015" - 2015, the 9th of May workdays: [Mon, Tue, Wed, Thu] weekend_eve: [Fri] weekend: [Sat, Sun] holidays: - "01.01" - "07.01" - "23.02" - "08.03" - "01.05" - "09.05" - "12.06" - "04.11"
Time classes divide the day of a certain category into time intervals. To specify the time class, you should create the time_classes/<time_class> parameter, as well as a parameters set for a time class that define the day category. Its values are represented by a list of time intervals. The default time_classes/default time class parameter is specified in a special way. If the time class can not be determined using the specified parameters, then the default time class will be used instead.
In the example below, the peak time class is specified and the time intervals for each day category considered to be peaking are defined. If no rule specifying the peak time class can be used, then the default time offpeak class will be used.
# # Time classes time_classes: # Defines the classes of time default: offpeak # Specifies the default value for time classes # All others items define the names of time classes and rules for them definition # Each rule is started from the name of day categorie and contains # the list of ranges of times in form "HH:MI:SS-HH:MI:SS" peak: workdays: - "00:00:00 - 01:00:00" - "17:00:00 - 23:59:59" weekend_eve: - "00:00:00 - 02:00:00" - "16:00:00 - 23:59:59" weekend: - "00:00:00 - 02:00:00" - "09:00:00 - 23:59:59" holidays: - "00:00:00 - 02:00:00" - "09:00:00 - 23:59:59"
The job parameters are specified as nested jobs parameter structures. There are the following tasks:
jobs/monitor - monitoring of the network interfaces and data cache usage;jobs/load - data cache loading/validating;jobs/scan - auxiliary task designed for data analysis Storages;
Further, when describing, the full parameter name will be based on the jobs parameter .
For the monitoring job, you should specify the incoming connection parameters (monitor/listener/host and monitor/listener/port), which will be used to receive messages from other processes. If one computer is used for all processes, then these parameters have to match the parameters defined for statistics collectors description.
In other words, the monitor/listener parameter specifies the server side of the IP connection, and the parameters specified within the statistics collectors description determine the client side one.
The monitoring job, in addition to gathering statistics of the processes operating with the cache, also gathers system information on network interfaces. In order to limit the list of network interfaces used to gather statistics, it is necessary to specify them in a list in the monitor/network_interfaces parameter.
monitor:
listener: # listener collects information from different processes about
# incoming/outgoing traffic
host: localhost
port: 1600
network_interfaces: # list of network interfaces which is used for getting
# incoming/outgoing traffic statistics
# if list is empty, all interfaces except "lo" will be used
-
Load parameters are divided into generic ones, which are specified as fields of the load parameter; used in offline-loading (load/offline); used in online-loading (load/online);
Generic load parameters:
# # Load jobs parameters load: ip_binding: # optional, list of local IP-adresses which could be used # to binding while loading - ignored_clients: # list of CIDRs for source IP addresses which should be ignored; # list could be defined internally into configuration file as list of CIDRs # under key "cidr_list" or into external files under key "cidr_files"; # if list is defined externally in file, each line that file must contain # only one CIDR cidr_list: - cidr_files: - rate_limits: # defines rate limit for data loading depending on time class; # value can be one of following: # - "unlimited" # - number of bytes which defines amount of loaded data per second offpeak: unlimited peak: 1m # ... skipping ...
The load/ip_binding parameter specifies the IP address list to be used when loading the data, the listed addresses are used evenly.
The load/ignored_clients parameter allows to specify the list of CIDR to be ignored, the requests being sent from these CIDR should not be taken into account when loading data. CIDRs can be specified both in the configuration file (the load/ignored_clients/cidr_list list), and in the external files (the load/ignored_clients/cidr_files list).
The load/rate_limits parameter defines the loading data speed limitations based on the time classes. To set a limit on the speed of loading data at a certain time, you should to create the load / rate_limits / <time class> parameter which value is used to set the desired limitation.
Offline-load parameters:
# # Load jobs parameters load: # ... skipping ... offline: parallel_workers: 1 # optional, number of parallel loading jobs, default 1 job_awaiting_time: 10 # awaiting time for job in a queue in seconds; # if after this time no one job is in the queue, # worker finishes its execution
Online-load parameters can be divided into the following logical groups:
Descriptions of IPFIX stream sources are specified in the load/online/exporters structure:
# # Load jobs parameters load: # ... skipping ... online: exporters: # list of IPFIX exporters main: queue_size: 1000 # optional, max number of messages from exporter which is sent to analyzing queue, # must be between 1 and 100000, default: 1000 host: localhost # host name or IP-address port: 1500 # listening port protocol: udp # one of possible protocols: UDP or TCP information_elements: # list of used information elements; # names are reserved, values must be in form "PEN/NUM" timestamp: "43823/1001" # mandatory, timestamp of event, must be IPFIX dateTimeSeconds host: "43823/1005" # mandatory, host name, must be IPFIX string path: "43823/1006" # mandatory, path to resource, must be IPFIX string login: "43823/1002" # optional, login, must be IPFIX string source_ip4: "43823/1003" # optional, source IP-address destination_ip4: "43823/1004" # optional, destination IP-address referal: "43823/1007" # optional, referal, must be IPFIX string user_agent: "43823/1008" # optional, user agent, must be IPFIX string cookie: "43823/1009" # optional, cookie, must be IPFIX string # ... skipping ...
To create a description of the IPFIX stream source, you should create the load/online/exporters/<IPFIX-источник>structure according to the example above. You should pay attention to specifying the following parameters:
load/online/exporters/<IPFIX-источник>/host and load/online/exporters/<IPFIX-источник>/port - definition of the server socket parameters (IPFIX stream should be sent to this address);load/online/exporters/<IPFIX-источник>/protocol - the definition of the protocol used to send messages;
The input IPFIX flow analysis description is specified in the load/online/analyzing structure:
# # Load jobs parameters load: # ... skipping ... online: # ... skipping ... analyzing: # the analyzing processes perform URL rewriting, filtering # and binding URLs getting from exporters to cache's definition # each analyzzng process sends valid URLs to collection queue parallel_workers: 1 # optional, number of parallel analyzing jobs, default 1 queue_size: 1000 # optional, max number of messages from each analyzing process # which is sent to collecting queue, must be between 1 and 100000, # default: 1000 # ... skipping ...
When specifying the analyzer parameters, you should focus initially on the load/online/analyzing/parallel_workers parameter. When there is a large number of messages from the processes receiving an IPFIX stream, the analyzer processes number should be increased. The indication to increase the number of analyzer processes is the message about the message queue overflow in the log files of the processes responsible for receiving the IPFIX stream.
The description of the requested objects collectors is specified in the load/online/collectors structure:
# # Load jobs parameters load: # ... skipping ... online: # ... skipping ... collectors: # list of collectors default: # the collector is used for aggregation of identical URLs # and distributes them into time windows # each event could hit no more than 1 window according to event's timestamp # the time window is a period of time which defines interval of time in the past # it is possible to have many time windows, in this case each window is # connected to previous one slots: 24 # optional, number of windows, could not be greater than 100, default 24 window: 3600 # optional, window size in seconds, could not be less than 60 secs, default 3600 week_by_4_hours: slots: 42 window: 14400 # ... skipping ...
The collector of the requested objects is the main factor for deciding whether to start the object download. Logically, the collector is a collection of counters in the sliding time intervals in the past from the present. When an object is requested, the counter is incremented in a certain interval, and as soon as the total value of all object counters reaches the threshold value specified within the cache description, the object loading is permitted.
In order to create the collector description, you should create the following parameters:
load/online/collectors/<object collector>/slots - number of intervals;load/online/collectors/<object collector>/window - the size of each interval;
In the example above, two collectors are described: default and week_by_4_hours. The default collector has 24 1-hour intervals, i.e. allows you to analyze the number of object requests during the day.
The description of the data loading processes is specified in the load/online/loading structure:
# # Load jobs parameters load: # ... skipping ... online: # ... skipping ... loading: parallel_workers: 1 # optional, number of parallel loading jobs, default 1 queue_size: 100 # optional, max number of messages from collecting process # which is sent to loading queue, must be between "parallel_workers" and 10000, # default: 100 unbuffered_queue_size: 10 # optional, if current queue size less or equal this value, the messages from # collecting process will put into loading queue immediately. The messages will be buffered # and ordered before puting into loading queue according to URLs' weight; # must be between 0 and "queue_size" - 2 * "parallel_workers", default 2 * "parallel_workers" # ... skipping ...
When specifying the load process parameters you should pay attention to the number of parallel processes. Do not significantly increase the queue size ( the load/online/loading/queue_size parameter), since when using a large value for this parameter at startup, the queue will be populated with objects that satisfy the load startup conditions, but it is possible that objects that are requested more frequently will appear during the queue processing.
Scanning is an auxiliary job that starts when loading and deleting data from the cache.
# # Scannig jobs parameters scan: workers: # the scanning processes perform scan cache directories on demand; # each scanning process sends information about found items to scanning queue parallel_workers: 4 # optional, number of parallel scanning jobs, default 4 job_queue_size: 5000 # optional, max number of messages in job queue, must be between 1000 and 10000 # default: 5000 result_queue_size: 100000 # optional, max number of messages from each scanner process # which are sent to result scanning queue, must be between 1000 and 1000000, # default: 100000
The functions performed by the scan task include the following:
Storage parameters are specified as the storage_parameters parameter fields. All storage parameters are divided into the following groups:
storage_parameters/general structurestorage_parameters/caches structureGeneral parameters specify the path to the storage root, as well as the maximum size that can be allocated to all the objects of the cache:
storage_parameters:
general:
path: "/var/cache/ccu/data" # optional, base path for caches, default: /var/cache/ccu/data;
max_size: "1T" # optional ("unlimited", 0 or missed - unlimited),
# maximum data size for all caches, possible suffixes are:
# Kb, Mb, Gb, Tb, Pb or without trailing "b" just K, M, G, T, P
The parameters of a particular cache are specified as the storage_parameters/caches/<cache description> structure, for example:
storage_parameters:
# ... skipping ...
caches:
youtube.com:
is_enabled: yes # optional (default yes), if value is "no" the cache definition
# will not be used for loading data
# ... skipping ...
specifies the cache description using the youtube.com name.
The parameter names will be based on the storage_parameters/caches/<cache description> parameter when describing cache parameters in this section.
The is_enabled parameter specifies whether objects will be loaded into the cache or not.
The cache statistics are specified in the statistics structure:
statistics:
group: "youtube.com" # name of statistic's group
# all statistics in the same group is aggregated together
collector: local # name of statistics collector
The statistics/group the parameter specifies the group where the cache operations will be taken account. The statistics/collector parameter specifies statistics collector, which will account for cache transactions.
For the online-loading, you need to specify the following parameters:
online:
collector: "week_by_4_hours" # name of collector which should be used
# for online processing
validating:
interval: 24h # revalidating interval for particular cache item;
# could be used only for caches with "general" algorithm
The online/collector parameter specifies the collector name of the requested objects. The online/validating/interval parameter can only be used by the cache with general load algorithm and specifies the time interval defining how often the cache-stored object will be checked against its original.
The following parameters are used to specify the load rules:
loading:
algorithm: "youtube.com" # internal algorithm name
required_weight: 3 # min URL's weight which is required to start loading, default 3
urls:
matching: # list of parameters for URL matching and rewriting;
# - the "key" item is used for unique identification of object;
# - the "target" item specifies target value for URL; usually target is used as
# as additional URL for loading
# - the "weight" item is used as weight addition, default 1;
# all items under "sources" are used as a source masks;
# the "key" and "target" can be omitted in this case
# the sources URLs will not be rewritten.
# it is necessary to have at least one URL in sources.
# It is possible to use in the key and target's expression variables from sources;
# See RE syntax on https://docs.python.org/2/library/re.html
- key: '\1'
weight: 1
sources:
- '^www\.youtube\.com/watch\?v=([a-zA-Z0-9_\-]+)'
- '^www\.youtube\.com/embed/([a-zA-Z0-9_\-]+)'
ignoring: # list of regular expr. for ignoring urls
- # default: all matched with matching parameters URLs will be processed
loadable_rejecting: # list of regular expr. for rejecting "real" urls
# which is obtained according to particular URL
# default: all "real" URLs will be processed
# See RE syntax on https://docs.python.org/2/library/re.html
- "itag=(?!(18|22|140)(?![0-9]))"
loading/algorithm parameter defines algorithm to be used for object loading. The following algorithms exist:loading/required_weight parameter specifies the required collector weight of the requested objects; when the weight is reached, object loading will start;loading/urls/matching parameter specifies the rules for original URL processing/overwriting; This parameter is a list of the elements structure with the following fields:key - the object key, which is issential for the loading algorithm, for example, the object key for the youtube.com algorithm is presented by the video identifier;weight - the weight to be added when the requested object in the collector will be accounted;target - rewritten original URL;sources - a list of regular expressions for analyzing original URL; key parameter value is the video id specified as the 1st group in regular expressions.loading/urls/ignoring parameter specifies a list of regular expressions for URLs that passed the matching stage ('loading/urls/matching') and to be excluded from the load;loading/urls/loadable_rejecting parameter specifies a list of regular expressions for loadable URLs to be excluded from the load; downloadable URLs are the links to real files obtained when the loading algorithm (loading/algorithm) is executed for original URL, for example, in order to execute the youtube.com algorithm for the URL specifying the video WEB-page to be viewed, a list of URLs for video and audio files will be received in all available formats.The following parameters are used to specify the rules for storing objects:
storage:
path: "sites/youtube.com" # regard to "storage_parameters/general/path"
levels: "2:2" # optional, store files in subfolders created by leading symbols
# of md5 sum of file name
# possible values:
# "1" - creating subfolder by the last symbol of md5 sum
# "2" - creating subfolder by the last two symbols of md5 sum
# "1:2" - creating subfolder by the last symbol of md5 sum
# and creating subfolder by 2th and 3th symbols
# from the end of md5 sum inside
# "2:2" - creating subfolder by the last two symbols of md5 sum
# and creating subfolder by 3th and 4th symbols
# from the end of md5 sum inside
max_size: "unlimited" # optional ("unlimited", 0 or missed - general limit will be used),
# maximum data size, possible suffixes are:
# Kb, Mb, Gb, Tb, Pb or without trailing "b" just K, M, G, T, P
expiry_time: "30d" # optional (0 or missed - unlimited), expiry time,
# default in seconds (without suffix)
# possible suffixes are:
# m - minutes
# h - hours
# d - days
storage/path parameter specifies the path for storing objects relative to storage_parameters/general/path parameter value;storage/levels parameter specifies the number of additional directories to be created based on md5-sum of the file name;storage/max_size parameter specifies the maximum total size of all objects stored in the cache. If this restriction is not specified, then the entire repository size restriction will be used instead;storage/expiry_time parameter specifies the lifetime of an object within the cache from the moment it is loaded;For cache objects, you can set constraints, for example:
constraints: # optional, specifies constraints for loaded files min_file_size: 128k # optional (default 0), min file size, possible suffixes are: # Kb, Mb, Gb, Tb, Pb or without trailing "b" just K, M, G, T, P max_file_size: "unlimited" # optional (default unlimited), max file size, possible suffixes are: # Kb, Mb, Gb, Tb, Pb or without trailing "b" just K, M, G, T, P # optional, command which should be executed after file loading; # if command returns non-zero result, the loaded file will be assumed as invalid an will be removed; # the next variables could be used in the command: # {cache_name} - cache name # {full_file_name} - fully qualified file name post_load_validation: "ft=`file -b {full_file_name}`; echo $ft; echo $ft | grep -E \"WebM|ISO Media|MPEG\" 2>/dev/null 1>&2"
It is possible to get caching of the the most frequently requested storage objects on an SSD, to do so you should define the following parameters:
# # SSD caching parameters ssd_caching: is_enabled: no # enable or not SSD caching, by default SSD caching is disabled path: "/var/cache/ccu/ssd" # optional, base path for SSD caching area, default: /var/cache/ccu/ssd # NOTES: # - to SSD caching you have to mount SSD on specified path (or create link /var/cache/ccu/ssd to mount point) # - changes this parameter should be performed in cooperation with changes in web server configuration # (it is recomended do not change this parameter) max_size: 128G # maximum data size for storing on SSD, possible suffixes are: # Kb, Mb, Gb, Tb, Pb or without trailing "b" just K, M, G, T, P required_weight: 10 # optional, min URL's weight which is required to start caching, default 10 uri_prefixes: # URI prefixes which are used in HTTP requests to access data ssd_cache_requests: "/ssd" # optional, prefix for requests to SSD cache, default: "/ssd" main_storage_requests: "/cache" # optional, prefix for requests to main storage, default: "/cache" frozen_time: 180 # optional, interval of time in seconds after putting file into cache in which that # file cannot be replaced by another one, default: 3 * collecting window # frozen time cannot be less than collecting window collector: # the collector is used for aggregation of identical requests # and distributes them into time windows slots: 60 # optional, number of windows, could not be greater than 120, default 60 window: 60 # optional, window size in seconds, could not be less than 60 secs, default 60 workers: # the worker processes perform scan cache directories on demand; # each process sends information about found items to result queue parallel_workers: 2 # optional, number of parallel jobs, default 2 job_queue_size: 5000 # optional, max number of messages in job queue, must be between 1000 and 10000 # default: 5000 result_queue_size: 100000 # optional, max number of messages from each process # which is sent to result queue, must be between 1000 and 1000000, # default: 100000
When setting parameters, you should pay attention to the number of working processes and the collector parameters.
When installing ccu the /etc/ccu/ccu.conf.default configuration template file will be created. To get the proper configuration, you should copy the template file to /etc/ccu/ccu.conf and edit/check the value of at least the following parameters: