EGI connectors

This document describes the available connectors for data in EGI infrastructure.


argo-egi-connectors is a bundle of connectors/sync components for various data sources established in EGI infrastructure, most notably GOCDB (EGI topology, downtimes), but there’s also support for fetching alternative EGI topology via various VO feeds, weights information via GStat service and POEM metric profiles.

Bundle consists of the following connectors:

  • component whose role is to filter out the messages coming from the argo-egi-consumer.

Connectors are syncing data on a daily basis. They are aware of the certain customer, associated jobs and their attributes and are generating and placing files into appropriate job folders. Data is written in a binary avro formated file which is suitable for processing at compute side. Topology, downtimes, weights and POEM profile information all together with a prefiltered metric results (status messages), represents an input for argo-compute-engine.


Installation narrows down to simply installing the package:

`yum -y install argo-egi-connectors`

Components require avro package to be installed/available.

File Types Destination
Configuration files /etc/argo-egi-connectors
Components /usr/libexec/argo-egi-connectors
Cronjobs (configured to be executed once per day) /etc/cron.d
Directory where components will put their files /var/lib/argo-connectors/EGI


Configuration of all components is centered around two configuration files: global.conf and customer.conf. Those files contains some shared config options and sections and are read by every component. There’s also a third one poem-connector.conf, specific only for because it needs some special treatment not available in first two’s.

Configuration file Description Shortcut
global.conf Config file is read by every component because every component needs to fetch host certificate to authenticate to a peer and to find correct avro schema. Description
customer.conf This configuration file lists all EGI jobs, their attributes and also all VOes and theirs set of jobs and attributes. Description
poem-connector.conf This configuration file is central configuration for Description


Config file is read by every component because every component needs to, at least, fetch host certificate to authenticate to a peer and to find correct avro schema. Additionally, some connectors have default sources of data specified that can be overidden in a next configuration file. Config options are case insensitive and whole config file is splitted into a few sections:

SchemaDir = /etc/argo-egi-connectors/schemas/

Every component generates output file in an avro binary format. This section points to a directory that holds all avro schemas.

VerifyServerCert = False
CAPAth = /etc/grid-security/certificates
HostKey = /etc/grid-security/hostkey.pem
HostCert = /etc/grid-security/hostcert.pem

Each component that talks to GOCDB or POEM peer authenticates itself with a host certificate. HostKey indicates the private and HostCert indicates the public part of certificate. Additionally, server certificate can be validated rounding up the mutual authentication. CAPath contains certificates of authorities from which chain will be tried to be built upon validating.

Downtimes = %(SchemaDir)s/downtimes.avsc
Poem = %(SchemaDir)s/metric_profiles.avsc
Prefilter = %(SchemaDir)s/metric_data.avsc
TopologyGroupOfEndpoints = %(SchemaDir)s/group_endpoints.avsc
TopologyGroupOfGroups = %(SchemaDir)s/group_groups.avsc
Weights = %(SchemaDir)s/weight_sites.avsc

This section, together with a [DEFAULT] section, constitutes the full path of avro schema file for each component. Avro schema files define the format of the data that each component is writing. Topology* schemas are common to and because there is a need of compute side to not make a difference between two topologies. Prefilter schema is taken from argo-egi-consumer since filters its metric results and needs to write them in the same format.

Downtimes = downtimes_%s.avro
Poem = poem_sync_%s.avro
Prefilter = prefilter_%s.avro
PrefilterConsumerFilePath = /var/lib/ar-consumer/ar-consumer_log_%s.avro
PrefilterPoem = poem_sync_%s.out
PrefilterPoemNameMapping = poem_name_mapping.cfg
TopologyGroupOfEndpoints = group_endpoints_%s.avro
TopologyGroupOfGroups = group_groups_%s.avro
Weights = weights_%s.avro

Section lists all the filenames that each component is generating. Directory is purposely omitted because it’s implicitly found in next configuration file. Exception is a PrefilterConsumerFilePath and PrefilterPoem options that tells the where to look for its input files. %s is a string placeholder that will be replaced by the date timestamp in format year_month_day.


This configuration file lists all customers, their jobs and appropriate attributes. Job is presented to argo-compute-engine as a folder with a set of files that are generated each day and that directs compute engine what metric results to take into account and do calculations upon them.

Directory structure

Job folders for each customer are placed under the customer’s OutputDir directory and appropriate directory names are read from the config file. Segment of configuration file that reflects the creation of directories is for example:

OutputDir = /var/lib/argo-connectors/Customer1
Jobs = JOB_Test1, JOB_Test2

Dirname = C1Testing1

Dirname = C2Testing2

OutputDir = /var/lib/argo-connectors/Customer2
Jobs = Job_Test3, JOB_Test4

Dirname = C2Testing1

Dirname = C2Testing2

This will result in the following jobs directories:


So there are two customers, C1 and C2, each one identified with its [CUSTOMER_*] section. CUSTOMER_ is a section keyword and must be specified when one wants to define a new customer. Each customer has two mandatory options: OutputDir and Jobs. With OutputDir option, customer defines his directory where he’ll write job folders and other data. Customer must also specify set of jobs listed in Jobs options since it can not exist without associated jobs. The name of the job folder is specified with Dirname option of the certain job so JOB\_Test1, identified with [JOB_Test1] section, will be named C1Testing1 and it will be placed under customer’s /var/lib/argo-connectors/Customer1/ directory. Each component will firstly try to create described directory structure if it doesn’t exist yet. Only afterward it will write its data.

Every connector reads this configuration file because it needs to find out how many customers are there and what are theirs customer and job directory names where they will lay down its files. So,,, all of them are writing theirs data in each job directory for each customer. Topology for EGI (fetched from GOCDB) is different than one for the VO so exceptions to this are and They are writing data for a job based on the job’s topology type specified with TopoType attribute.

Job attributes

Besides Dirname option that is common for all connectors, some of them have job attributes that are relevant only for them and upon which they are changing their behaviour. Some of those attributes are mandatory like Profiles and TopoType and the other ones like TopoSelect* attributes are optional. Furthermore, as there are two kind of topologies, there are also two set of job attributes and values.

GOCDB topology
Dirname = EGI_Cloudmon
Profiles = CLOUD-MON
TopoType = GOCDB
TopoFeed =
TopoFetchType = ServiceGroups
TopoSelectGroupOfEndpoints = Monitored:Y, Scope:EGI, Production:Y
TopoSelectGroupOfGroups = Monitored:Y, Scope:EGI

This is an example of the job that fetchs topology from GOCDB since TopoType attribute is set to GOCDB. Profiles is an attribute relevant to so for this job will write CLOUD-MON profile in EGI_Cloudmon job folder under /EGI directory. Topo* attributes are relevant for TopoFeed attribute in the context of the GOCDB topology is optional. If it’s specified, it will override default source of topology which is

Topology is separated in two abstracts:

  • group of groups
  • group of service endpoints

Service endpoints are grouped either by the means of Sites or Service groups. Those are listed and represented as an upper level abstract of group of service endpoints - group of groups. Customer can fetch either Sites and their corresponding endpoints or Service groups and their corresponding endpoints per job, but not both of them. What is being fetched is specified with TopoFetchType option/job attribute. For each abstract there will be written TopologyGroupOfGroups and TopologyGroupOfEndpoints filenames (specified in global.conf) into appropriate job folder. TopoSelectGroupOfGroups and TopoSelectGroupOfEndpoints options are used for further filtering. Values are set of tags used for picking up matching entity existing in the given abstract. Tags for group of groups are different for Sites and Service groups. In contrary, set of tags for groups of endpoints remains the same no matter what type of fetch customer specified.

So, in a TopoFetchType option customer can either specify:

  • ServiceGroups - to fetch Service groups
  • Sites - to fetch Sites

Tags represent a fine-grained control of what is being written in output files. It’s a convenient way of selecting only certain entities, being it Sites, Service groups or Service endpoints based on appropriate criteria. Tags are optional so if a certain tag for a corresponding entity is omitted, than filtering is not done. In that case, it can be considered that entity is fetched for all its values of an omitted tag.

Group of group tags are different for a different type of fetch. Tags and values for a different entities are:


  • Certification = {Certified, Uncertified, Closed, Suspended, Candidate}
  • Infrastructure = {Production, Test}
  • Scope = {EGI, Local}


  • Monitored = {Y, N}
  • Scope = {EGI, Local}

Tags for selecting group of endpoints are:

Service Endpoints

  • Production = {Y, N}
  • Monitored = {Y, N}
  • Scope = {EGI, Local}
VO topology
BioMed =

Dirname = BioMed_Critical
TopoType = VOFeed
TopoFeed = %(BioMed)s
TopoSelectGroupOfGroups = Type:(OPS_Tier, OPS_Site)

This is an example of the job that is fetching topology from provided VO feed since TopoType attribute is set to VOFeed. Again, Profiles attribute is mandatory and is relevant to which will write ROC_CRITICAL profile in BioMed_Critical job folder. Topo* attributes are relevant for Contrary to GOCDB topology jobs, TopoFeed attribute for jobs dealing with the VO topology is mandatory and must be specified. Although same topology feed can be specified across multiple jobs, internally, data from one feed is fetched only once and is further filtered and written for every job.

VO topology is also separated and written in two abstracts, group of groups and group of service endpoints, but there are no tags since VO itself filters and handles what sites and service endpoints to take into account and defines the VO groups they belong to. With that being said, there is a TopoSelectGroupOfGroups option available which is used to pick up VO groups based on their type. VO groups are entities existing in the group of group abstract of topology. In the example above, will pick up VO groups that match OPS_Site and OPS_Tier types and write them into TopologyGroupOfGroups file. Endpoints are written in TopologyGroupOfEndpoints file.

Data feeds

Source of the data for other connectors like and are optional and can be specified per job. If specified, they will override their default source of data. Example:

Dirname = BioMed_Critical
TopoType = VOFeed
TopoFeed = %(BioMed)s
TopoSelectGroupOfGroups = Type:(OPS_Tier, OPS_Site)
WeightsFeed = 
DowntimesFeed =

WeightsFeed and DowntimesFeed are alternative data feeds for this job for connectors and, respectively.


This configuration file is central configuration for whose role is:

  • fetch all defined POEM profiles from each POEM server specified
  • prepare and layout data needed for

POEM profiles fetch

Config file is splitted into a few sections:

Host =
VO = ops

This section defines the URL where POEM server is located and all VOes for which POEM profiles will be fetched. Multiple POEM servers can be specified by defining multiple POEM server sections:

Host = poem1
VO = vo1, vo2

Host = poem2
VO = vo3, vo4

Same POEM profile can be defined on multiple POEM servers. Each POEM server can further extend it with a custom combinations of metrics and service flavours. To distinguish POEM profile defined on multiple POEM servers, namespace must be used. One must be aware of the namespace that POEM server exposes and specify it in FetchProfiles section:


Prefilter data also generates plaintext PrefilterPoem file (specified in global.conf) on a daily basis for each customer and places it under customer’s directory. Content of the file is controlled in [PrefilterData] section:

AllowedNGI =
AllowedNGIProfiles =,,,
AllNGI1 =
AllNGIProfiles1 =,
AllNGI2 =
AllNGIProfiles2 =

AllowedNGI option defines remote config file that states all allowed NGIes and corresponding nagios boxes. All of them will be expanded and listed together with the information from AllowedNGIProfiles POEM profiles (metrics, service flavours, VOes).

AllNGI1 option is similar in sense that it will extended specified nagios box (monitoring instance) with the information from AllNGIProfiles1 POEM profiles. Multiple AllNGI* options can be specified and they must come in pair fashion so for every AllNGI[n] option, there must exist AllNGIProfiles[n] option that is related to it.

With all these informations written in PrefilterPoem file, can do its work, so it will filter consumer messages if:

  • message that enter into broker network doesn’t come from allowed NGI or nagios box for certain NGI is incorrectly specified
  • metric result is response to metric not found in a fetched service flavour
  • metric result’s service flavour is not registered in any fetched POEM profile
  • metric result is registered for different VO, not specified in VO option of any [PoemServer] section





	BioMed =

	OutputDir = /var/lib/argo-connectors/EGI/

	Jobs = JOB_EGICritical, JOB_EGICloudmon, JOB_BioMedCloudmon, JOB_BioMedCritical

	Dirname = EGI_Critical
	Profiles = ROC_CRITICAL
	TopoType = GOCDB
	TopoFetchType = Sites
	#TopoSelectGroupOfEndpoints = Production:Y, Monitored:Y, Scope:EGI
	TopoSelectGroupOfGroups = Certification:Uncertified, Infrastructure:Test, Scope:EGI

	Dirname = EGI_Cloudmon
	Profiles = CLOUD-MON
	TopoType = GOCDB
	TopoFetchType = ServiceGroups
	TopoSelectGroupOfEndpoints = Monitored:Y, Scope:EGI, Production:N
	#TopoSelectGroupOfGroups = Monitored:Y, Scope:EGI, Certification:(Certified,Candidate)

	Dirname = BioMed_Critical
	Profiles = ROC_CRITICAL
	TopoType = VOFeed
	TopoFeed = %(BioMed)s
	TopoSelectGroupOfGroups = Type:OPS_Site

	Dirname = BioMed_Cloudmon
	Profiles = CLOUD-MON
	TopoType = VOFeed
	TopoFeed = %(BioMed)s
	#TopoSelectGroupOfGroups = Type:OPS_Tier


Customer jobs:




Prefilter data:




For customer's job JOB_EGICritical, we are selecting only those sites that match `Certification:Uncertified`, `Infrastructure:Test` and `Scope:EGI`, so in `TopologyGroupOfGroups` file there will be only those sites listed:


	 % avro cat /var/lib/argo-connectors/EGI/EGI_Critical/group_groups_2015_04_07.avro | tail -n 1
	 {"group": "Russia", "tags": {"scope": "EGI", "infrastructure": "Test", "certification": "Uncertified"}, "type": "NGI", "subgroup": "SU-Protvino-IHEP"}



For customer's JOB_EGICloudmon, we are selecting only those service endpoints that match `Monitored:Y`, `Scope:EGI`, `Production:N`:


	 % avro cat /var/lib/argo-connectors/EGI/EGI_Cloudmon/group_endpoints_2015_04_07.avro
	 {"group": "ROC_RU_SERVICE", "hostname": "", "type": "SERVICEGROUPS", "service": "Top-BDII", "tags": {"scope": "EGI", "production": 0, "monitored": 1}}



JOB_BioMedCritical is taking into account only OPS\_Site VO groups:


	% avro cat /var/lib/argo-connectors/EGI/BioMed_Critical/group_groups_2015_04_07.avro | tail -n 5
	{"group": "SAMPA", "tags": null, "type": "OPS_Site", "subgroup": "SAMPA"}
	{"group": "UPJS-Kosice", "tags": null, "type": "OPS_Site", "subgroup": "UPJS-Kosice"}
	{"group": "GR-06-IASA", "tags": null, "type": "OPS_Site", "subgroup": "GR-06-IASA"}
	{"group": "FI_HIP_T2", "tags": null, "type": "OPS_Site", "subgroup": "FI_HIP_T2"}
	{"group": "UKI-SOUTHGRID-RALPP", "tags": null, "type": "OPS_Site", "subgroup": "UKI-SOUTHGRID-RALPP"}



JOB_BioMedCloudmon requires only CLOUD-MON POEM profile so in `Poem` file you have:


	 % avro cat  /var/lib/argo-connectors/EGI/BioMed_Cloudmon/poem_sync_2015_04_07.avro | tail -n 5
	 {"profile": "", "metric": "", "service": "egi.Perun", "tags": {"fqan": "", "vo": "ops"}}
	 {"profile": "", "metric": "", "service": "", "tags": {"fqan": "", "vo": "ops"}}
	 {"profile": "", "metric": "org.nagios.Broker-TCP", "service": "", "tags": {"fqan": "", "vo": "ops"}}
	 {"profile": "", "metric": "org.nagios.Broker-TCP", "service": "", "tags": {"fqan": "", "vo": "ops"}}
	 {"profile": "", "metric": "org.nagios.Broker-TCP", "service": "", "tags": {"fqan": "", "vo": "ops"}}





	% /usr/libexec/argo-egi-connectors/ -d 2015-04-07
	% find /var/lib/argo-connectors -name '*downtimes*'

Connectors are using following GOCDB PI methods: is using POEM PI method: - POEM - metrics_in_profiles

Construction of VO feeds

GRNET S.A - Greek Research & Technology Network Centre national de la recherche scientifique SCRE - Web stranice Sveučilišnog računskog centra European Grid Infrastructure