Overview

In the scope of PERF-524 it's needed to run tests to answer questions:

Can the current System accommodate an average load? Load model described here NLA load model investigation and creation.
- If not PERF-525
What happened at peak times when all workflows are running at once?
Typical KPIs:
- Service CPU
- Service Memory
- DB CPU
- DB Memory
Response times
Duration of long workflows
Recommendations to improve on scaling up/out modules to accommodate peak times

Summary

The current system can accommodate an average load only without Data Import. With data import, we will have 'HTTP 500 Internal Server Error. If the issue persists, please report it to EBSCO Connect.' for several workflows in the source-storage/records/{id}/formatted requests PERF-582 - Getting issue details... STATUS , and the general response time will be longer up to 2 times for all other workflows.
According to the ticket PERF-525 - Getting issue details... STATUS , were performed a set of tests with changing parameters of DI modules. Results in the table show that for the configuration of mod-source-record-manager with 500 database connections count and mod-source-record-storage with 500 database connections count and 60-sec timeout it was not observed "HTTP 500 Internal Server Error." issue, but the response time for requests like GET /source-storage/records/id/formatted greatly increases during DI (up to 75 sec).
Increasing the number of database connections to DB_MAXPOOLSIZE = 200 for mod-source-record-manager and mod-source-record-storage did not give any positive effect on 'HTTP 500 Internal Server Error. If the issue persists, please report it to EBSCO Connect.' but duration of Data Import job was decreased twice.
Service CPU utilization did not exceed 36% even at the beginning of all processes together. Without spikes at the beginning and start of DI jobs average CPU usage was about 15%. Instance CPU Utilization did not exceed 22%.
Service memory utilization was stable, and no memory leaks were suspected during tests.
Each FYR (Fiscal close - end of FY rollover) job is consuming a lot of DB CPU (each spike here corresponds to each FYR job). Approximately DB CPU usage was up to 92%

Recommendations & Jiras (Optional)

Jiras

PERF-582 - Getting issue details... STATUS

Test Runs & Results

Test #	# configuration	Test duration	comments
1	All workflows started at the same time	1 hour	Res 500 ERROR during Data Import (DI) process
2	Data import and data export started with 15 min delay	1 hour	Res 500 ERROR during Data Import process
3	Data import and data export started with 1 min delay	1 hour	Res 500 ERROR during Data Import process
4	Data import started with 1 min delay and data export started with 20 min delay	35 min	Res 500 ERROR during Data Import process
5	Same as Test# 4 Jenkins configuration with increased DB_MAXPOOLSIZE = 200 for mod_srs and mod_srm	30 min	Res 500 ERROR during Data Import process. Data import duration decreased 2 times.
6	Test without Data import	1 hour	NO errors

Test results from 1st test run (1st, 2nd and 3rd test run results are similar):

Test #	Workflow name	Total time it takes to complete workflow				After environment improvement, no server errors were observed (23-06-2023)***		Time-consuming requests for each workflow during DI for default configuration, finished with Response body: HTTP 500 Internal Server Error. If the issue persists, please report it to EBSCO Connect.
		Avg With DI (sec)	95th pct with DI (sec)	Avg no DI (sec)	95th pct no DI (sec)	With DI Avg	Without DI
1	CICO_Checkin	1.238	1.506	1.054	1.591	1.188	1.089
2	CICO_Checkout	2.156	2.829	1.650	1.948	2.040	1.933
3	IO_View invoices	0.907	1.305	0.763	0.913	1.178	1.015
4	IO_Create invoices	1.433	1.815	1.174	1.370	1.689	1.795
5	IO_Edit invoices	1.983	2.422	1.581	1.897	2.276	1.015
6	IO_Delete invoices	1.070	1.196	0.804	0.927	1.248	1.175
7	AIE_Approving Invoices	1.752	2.211	1.453	1.940	3.208	2.740
8	VAR_View Authority records	22.037	30.604	0.289	0.381	49.268	0.295	VAR_GET /source-storage/records/marc_id/formatted
9	VTT_View MARC tag table	41.272	61.935	0.987	1.284	75.006	1.052	VTT_GET source-storage/records/{id}/formatted *2
10	VH_View holdings records	27.328	33.579	1.526	1.922	34.645	1.291	VH_GET source-storage/records/{id}/formatted
11	VB_View Bib	22.851	31.634	0.841	1.168	42.175	0.900	VB_GET source-storage/records/{id}/formatted
12	PRO_View patron records	0.672	1.118	0.566	0.883	0.359	0.638
13	PRO_Delete patron records	0.892	1.336	0.638	1.070	0.728	0.763
14	PRO_Update patron records	1.386	2.097	1.043	1.625	1.166	1.157
15	PRO_Create patron records	1.547	1.979	1.098	1.261	1.571	1.286
16	LO_View Ledger	0.122	0.458	0.050	0.088	0.125	0.047
17	LO_Create ledger	0.684	0.840	0.616	0.761	0.542	0.629
18	LO_Edit ledger	0.076	0.094	0.054	0.085	0.698	0.047
19	LO_Delete a ledger	0.080	0.129	0.046	0.080	0.073	0.050
20	DE_Export bib "Default instances export job profile"	11 sec (5000 records)	-	5 sec (5000 records)	-	6 sec (5000 records)	5 sec (5000 records)
21	DE_Export holdings "Default holdings export job profile"	3 min 16 sec (5000 records)	-	26 sec (5000 records)	-	28 sec (5000 records)	27 sec (5000 records)
22	DE_Export authority records "Default authority export job profile"	8 sec (5000 records)	-	3 sec (5000 records)	-	3 sec (5000 records)	3 sec (5000 records)
23	DI "DISC HRID match"	1sec (1 record)	-	-	-	1 sec	-
24	DI "DS LA edeposit records update"	17 min 5 sec	-	-	-	3 min 9 sec	-
25	DI "DISC New edeposit records"	-	-	-	-	13 sec	-
26	DI "DISC New NON edeposit records"	4 sec (5 records)	-	-	-	3 sec	-
27	IRO_View item records	24.771	32.327	1.289	1.649	37.804	1.449	IRO_GET source-storage/records/{id}/formatted
28	IRO_update item records	17.226	33.950	0.998	1.250	-	1.033	IRO_GET source-storage/records/{id}/formatted
29	IRO_delete item records	23.686	31.497	0.927	1.099	43.122	1.042	IRO_GET source-storage/records/{id}/formatted
30	MPS_Monitoring Pick Slips and Requests GET /circulation/requests	0.434	0.527	0.359	0.480	0.324	0.384
31	MPS_Monitoring Pick Slips and Requests GET /circulation/pick-slips/	0.106	0.312	0.112	0.256	0.086	0.116
32	MPS_Monitoring Pick Slips and Requests	0.297	0.297	0.303	0.303	0.136	0.184
33	ULR_Users loan renewal	1.899	2.395	1.467	1.661	1.939	1.564
34	ILR_Item-level requests	0.835	1.267	0.669	0.973	0.722	0.717
35	VRO_View vendor records	1.315	2.305	0.713	1.165	0.379	0.773
36	VRO_Edit vendor records	7.429	9.980	5.199	6.190	8.109	6.036
37	VRO_Create vendor records	1.436	1.829	1.064	1.200	1.579	1.189
38	VRO_Delete vendor records	0.635	0.885	0.412	0.522	0.266	0.458
39	POO_Create purchase orders	1.974	2.393	1.625	1.733	1.726	1.840
40	POO_View purchase orders	1.502	1.603	1.205	1.435	0.809	1.006
41	POO_Edit purchase orders	3.118	3.914	2.076	2.984	1.095	1.829
42	POO_Delete purchase orders	2.036	3.387	1.432	1.830	0.779	1.110
43	RIH_Retrieving instances and holdings	19.737	30.084	0.035	0.073	12.515	0.039	RIH_GET source-storage/source-records
44	ETT_Edit MARC tag table	118.391	128.508	3.424	4.257	87.175	3.599	ETT_GET /source-storage/records/instance_id/formatted ETT_GET /records-editor/records
45	FYR_Fiscal close - end of FY rollover	10 min 30 s	-	11 min	-	-	13 min
46	VIR_Blacklight: View an inventory record JMeter script	1.030	1.602	0.821	1.042	0.934	0.839
47	BLS_Blacklight: Create a Request JMeter script	1.336	1.605	1.122	1.404	1.139	1.156
48	PRV_Blacklight: Create a View Patron record JMeter script	0.106	0.136	0.073	0.110	0.078	0.073
49	VIH_View instance holdings details	21.241	31.698	1.456	1.572	47.835	1.535	VIH_GET /source-storage/records/instanceId/formatted

*Note that workflows that have response times or durations in red are the ones that are at least 2 times higher than when running without Data Import jobs.
*** Test run results with the configuration of mod-source-record-manager with 500 database connections count and mod-source-record-storage with 500 database connections count and 60-sec timeout.

Throughput graphs

For Test#1, Test#3 and Test#4 graphs are almost the same.

Memory Utilization

Memory utilization of the most memory-consuming modules & DI modules:

mod-inventory - 98%
mod-orders - 73%
mod-source-record-manager - 44%
mod-source-record-storage - 23,5%
mod-inventory-storage - 31%
data-import - 22%
mod-di-converter-storage - 31%

This graph represents memory usage of 3 first test runs and shows that no memory leak is suspected for all of the modules.

Service CPU Utilization

CPU usage did not exceed 36 % for all modules. We can observe spikes in CPU usage of DI modules at the beginning of the Data Import jobs. Without DI spikes average CPU usage was about 15%.

Test# 1 - Test# 4 DI duration - 17 min.
Test# 5 DI duration - 8 min.

Most CPU-consuming modules:

mod-inventory - 36%
mod-authtoken -25%
mod-data-import - 16%
mod-di-converter-storage -15,6%
mod-quick-marc - 15%
mod-finance-storage - 14%
nginx-okapi - 14%
okapi - 11%
mod-tags -10,6%
others - usage less than 10%
mod-source-record-manager - up to 7%
mod-source-record-storage - up to 4%

Instance CPU Utilization

RDS CPU Utilization

Each FYR (Fiscal close - end of FY rollover) job is consuming a lot of DB CPU (each spike here corresponds to each FYR job).

Approximately DB CPU usage is up to 92%

RDS Database Connections
Test# 1 - Test# 4 for part of the test with DI job- 620 connections count.
Test# 5 for part of the test with DI job- 820 connections count. (allocated additional 200 connections for DI modules)

Appendix

Infrastructure

Records count :

mod_source_record_storage.marc_records_lb = 7300919
mod_source_record_storage.raw_records_lb = 7300919
mod_source_record_storage.records_lb = 7300919
mod_source_record_storage.marc_indexers = 245032159 (all records)
mod_source_record_storage.marc_indexers with field_no 010 = 1008129
mod_source_record_storage.marc_indexers with field_no 035 = 8968420
mod_inventory_storage.authority = 852215
mod_inventory_storage.holdings_record = 6091403
mod_inventory_storage.instance = 5581816
mod_inventory_storage.item = 5705915

PTF -environment ncp3

9 m6i.2xlarge EC2 instances located in US East (N. Virginia)us-east-1
2 database instances, one reader, and one writer
Name API Name Memory GIB vCPUs max_connections
R6G Extra Large db.r6g.xlarge 32 GiB 4 vCPUs 2731
MSK ptf-kakfa-3
- 4 m5.2xlarge brokers in 2 zones
- Apache Kafka version 2.8.0
- EBS storage volume per broker 300 GiB
- auto.create.topics.enable=true
- log.retention.minutes=480
- default.replication.factor=3
Kafka topics partitioning: - 2 partitions for DI topics

Modules memory and CPU parameters

Modules	Version	Task Definition	Running Tasks	CPU	Memory	MemoryReservation	MaxMetaspaceSize	Xmx
mod-inventory-storage	26.0.0	1	2	1024	2208	1952	384	1440
mod-inventory	20.0.4	1	2	1024	2880	2592	512	1814
mod-tags	2.0.1	1	2	128	1024	896	128	768
mod-gobi	2.6.0	1	2	128	1024	896	128	700
mod-remote-storage	2.0.2	1	2	1024	4920	4472	512	3960
mod-invoice-storage	5.6.0	1	2	128	1024	896	128	700
edge-sip2	3.0.0	1	2	128	1024	896	128	768
mod-users-bl	7.5.0	1	2	512	1440	1152	128	922
edge-rtac	2.6.0	1	2	128	1024	896	128	768
mod-feesfines	18.2.1	1	2	128	1024	896	128	768
mod-rtac	3.5.0	1	2	128	1024	896	128	768
mod-erm-usage-harvester	4.3.0	1	2	128	1024	896	128	768
mod-search	2.0.1	1	2	400	2592	2480	1024	1440
mod-service-interaction	2.2.2	1	2	256	2048	1844	512	1290
edge-ncip	1.8.1	1	2	128	1024	896	128	768
mod-authtoken	2.13.0	1	2	512	1440	1152	128	922
mod-permissions	6.3.1	2	2	512	1684	1544	512	1024
mod-circulation-storage	16.0.0	1	2	1024	1536	1440	512	896
mod-ncip	1.13.1	1	2	128	1024	896	128	768
mod-pubsub	2.9.1	1	2	1024	1536	1440	512	922
edge-orders	2.8.1	1	2	1024	1536	1440	512	922
mod-circulation	23.5.4	1	2	1536	2880	2592	128	700
edge-caiasoft	2.0.0	1	2	128	1024	896	-	-
mod-data-export	4.7.1	1	1	1024	1024	896	128	768
mod-organizations-storage	4.5.1	1	2	128	1024	896	128	700
mod-source-record-storage	5.6.5	1	2	2048	5600	5000	512	3600
mod-copycat	1.4.0	1	2	896	1024	896	128	768
mod-bulk-operations	1.0.5	1	2	1024	3072	2600	512	1536
mod-quick-marc	3.0.0	1	1	128	2288	2176	512	1664
mod-audit	2.7.0	1	2	1024	1024	896	128	768
mod-oai-pmh	3.11.3	1	2	1024	2248	2000	512	1440
edge-connexion	1.0.6	1	2	128	1024	896	128	768
mod-kb-ebsco-java	3.13.0	1	2	128	1024	896	128	768
mod-patron	5.5.2	1	2	128	1024	896	128	768
mod-email	1.15.3	1	2	128	1024	896	128	768
mod-password-validator	3.0.0	1	2	128	1440	1298	512	768
mod-login	7.9.0	1	2	1024	1440	1298	512	768
mod-data-export-worker	3.0.12	1	2	1024	3072	2600	512	2048
mod-agreements	5.5.2	1	2	128	3096	2580	512	2048
edge-oai-pmh	2.6.1	1	2	1024	1512	1360	512	1440
mod-eusage-reports	1.3.0	1	2	128	1024	896	128	768
mod-orders-storage	13.5.0	1	2	512	1024	896	128	700
mod-notify	3.0.0	1	2	128	1024	896	128	768
mod-source-record-manager	3.6.2	1	2	2048	5600	5000	512	3600
mod-di-converter-storage	2.0.2	2	2	128	1024	896	128	768
mod-template-engine	1.18.0	1	2	128	1024	896	128	768
mod-user-import	3.7.2	1	2	128	1024	896	128	768
mod-finance-storage	8.4.1	1	2	128	1024	896	128	700
mod-users	19.1.1	1	2	128	1024	896	128	768
mod-sender	1.10.0	1	2	128	1024	896	128	768
mod-graphql	1.11.0	1	2	128	1024	896	128	768
mod-licenses	4.3.1	1	2	128	2480	2312	512	1792
mod-invoice-b	5.6.2	1	2	512	1440	1152	128	922
mod-event-config	2.5.0	1	2	128	1024	896	128	768
mod-calendar	2.4.2	1	2	128	1024	896	128	768
mod-erm-usage	4.5.2	1	2	128	1024	896	128	768
mod-patron-blocks	1.8.0	1	2	1024	1024	896	128	768
mod-data-import	2.7.1	1	1	256	2048	1844	512	1292
mod-ebsconet	2.0.0	1	2	128	1248	1024	256	700
edge-dematic	2.0.0	1	2	128	1024	896	-	-
mod-task-list	5.0.1	1	1	128	1024	896	128	768
mod-courses	1.4.7	1	2	128	1024	896	128	768
mod-inventory-update	3.0.1	1	2	128	1024	896	128	768
mod-login-saml	2.6.1	1	2	128	1024	896	128	768
mod-orders	12.6.6	1	2	1024	2048	1440 (Recommended to change to 1544)	512	1024
mod-configuration	5.9.1	1	2	128	1024	896	128	768
mod-organizations	1.7.0	1	2	128	1024	896	128	700
mod-notes	5.0.1	1	2	128	1024	896	128	322
mod-finance	4.7.1	1	2	128	1024	896	128	700
mod-data-export-spring	2.0.1	1	1	256	2048	1844	256	1292
edge-patron	4.11.0	1	2	256	1024	896	128	768
okapi	5.0.1	2	3	1024	1684	1440	512	922
nginx-okapi	2022.03.02	1	2	128	1024	896	-	-
pub-okapi	2022.03.02	1	2	128	1024	896	-	768

Methodology/Approach

To test Baseline for normal NLA library usage the JMeter scripts were used.

Tested with different DI delays:

From test start
1 min delay
20 min delay
without DI

Data was gathered from 2 periods with and without data import.

DI - data import
FYR - Fiscal close - end of FY rollover

Folio Development Teams

Master Script normal load test - NLA report

Overview

Summary

Recommendations & Jiras (Optional)

Jiras

Test Runs & Results

Throughput graphs

Memory Utilization

Service CPU Utilization

Instance CPU Utilization

RDS CPU Utilization

RDS Database Connections
Test# 1 - Test# 4 for part of the test with DI job- 620 connections count.
Test# 5 for part of the test with DI job- 820 connections count. (allocated additional 200 connections for DI modules)

Appendix

Infrastructure

Methodology/Approach

Name	API Name	Memory GIB	vCPUs	max_connections
R6G Extra Large	db.r6g.xlarge	32 GiB	4 vCPUs	2731

Master Script normal load test - NLA report

Overview

Summary

Recommendations & Jiras (Optional)

Jiras

Test Runs & Results

Throughput graphs

Memory Utilization

Service CPU Utilization

Instance CPU Utilization

RDS CPU Utilization

RDS Database ConnectionsTest# 1 - Test# 4 for part of the test with DI job- 620 connections count.Test# 5 for part of the test with DI job- 820 connections count. (allocated additional 200 connections for DI modules)

Appendix

Infrastructure

Methodology/Approach

RDS Database Connections
Test# 1 - Test# 4 for part of the test with DI job- 620 connections count.
Test# 5 for part of the test with DI job- 820 connections count. (allocated additional 200 connections for DI modules)