WAN Latency: Rsync versus SCP

We were curious about what methods we can use to manage large files that must be copied between sites with WAN-type latency and also restrict ourselves to processes available on the CA Identity Suite virtual appliance / Symantec IGA solution.

Leveraging VMware Workstation’s ability to introduce network latency between images, allows for a validation of a global password reset solution.

If we experience deployment challenges with native copy operations, we need to ensure we have alternatives to address any out-of-sync data.

The embedded CA Directory maintains the data tier in separate binary files, using a software router to join the data tier into a virtual directory. This allows for scalability and growth to accommodate the largest of sites.

We focused on the provisioning directory (IMPD) as our likely candidate for re-syncing.

Test Conditions:

  1. To ensure the data was being securely copied, we kept the requirement for SSH sessions between two (2) different nodes of a cluster.
  2. We introduce latency with VMware Workstation NIC for one of the nodes.

3. The four (4) IMPD Data DSAs were resized to 2500 MB each (a similar size we have seen in production at many sites).

4. We removed data and the folder structure from the receiving node to avoid any checksum restart processes from gaining an unfair advantage.

5. If the process allowed for exclusions, we did take advantage of this feature.

6. The feature/process/commands must be available on the vApp to the ‘config’ or ‘dsa’ userIDs.

7. The reference host/node that is being pulled, has the CA Directory Data DSAs offline (dxserver stop all) to prevent ongoing changes to the files during the copy operation.

Observations:

SCP without Compression: Unable to exclude other files (*.tx,*.dp, UserStore) – This process took over 12 minutes to copy 10,250 MB of data

SCP with Compression: Unable to exclude other files (*.tx,*.dp, UserStore) – This process still took over 12 minutes to copy 10,250 MB of data

Rsync without compression: This process can exclude files/folders and has built-in checksum features (to allow a restart of a file if the connection is broken) and works over SSH as well. If the folder was not deleted prior, then this process would give artificial high-speed results. This process was able to exclude the UserStore DSA files and the transaction files (*.dp & *.tx) that are not required to be copied for use on a remote server. Only 10,000 MB (4 x 2500 MB) was copied instead of an extra 250 MB.

Rsync with compression: This process can exclude files/folders and has built-in checksum features (to allow a restart of a file if the connection is broken) and works over SSH as well. This process was the winner, and; extremely amazing performance over the other processes.

Total Time: 1 min 10 seconds for 10,000 MB of data over a WAN latency of 70 ms (140 ms R/T)

Now that we have found our winner, we need to do a few post steps to use the copied files. CA Directory, to maintain uniqueness between peer members of the multi-write (MW) group, have a unique name for the data folder and the data file. On the CA Identity Suite / Symantec IGA Virtual Appliance, pseudo nomenclature is used with two (2) digits.

The next step is to rename the folder and the files. Since the vApp is locked down for installing other tools that may be available for rename operations, we utilized the find and mv command with a regular xpression process to assist with these two (2) steps.

Complete Process Summarized with Validation

The below process was written within the default shell of ‘dsa’ userID ‘csh’. If the shell is changed to ‘bash’; update accordingly.

The below process also utilized a SSH RSA private/public key process that was previously generated for the ‘dsa’ user ID. If you are using the vApp, change the userID to config; and su – dsa to complete the necessary steps. You may need to add a copy operation between dsa & config userIDs.

Summary of using rsync with find/mv to rename copied IMPD *.db files/folders
[dsa@pwdha03 ~/data]$ dxserver status
ca-prov-srv-03-impd-main started
ca-prov-srv-03-impd-notify started
ca-prov-srv-03-impd-co started
ca-prov-srv-03-impd-inc started
ca-prov-srv-03-imps-router started
[dsa@pwdha03 ~/data]$ dxserver stop all > & /dev/null
[dsa@pwdha03 ~/data]$ du -hs
9.4G    .
[dsa@pwdha03 ~/data]$ eval `ssh-agent` && ssh-add
Agent pid 5395
Enter passphrase for /opt/CA/Directory/dxserver/.ssh/id_rsa:
Identity added: /opt/CA/Directory/dxserver/.ssh/id_rsa (/opt/CA/Directory/dxserver/.ssh/id_rsa)
[dsa@pwdha03 ~/data]$ rm -rf *
[dsa@pwdha03 ~/data]$ du -hs
4.0K    .
[dsa@pwdha03 ~/data]$ time rsync --progress -e 'ssh -ax' -avz --exclude "User*" --exclude "*.dp" --exclude "*.tx" dsa@192.168.242.135:./data/ $DXHOME/data
FIPS mode initialized
receiving incremental file list
./
ca-prov-srv-01-impd-co/
ca-prov-srv-01-impd-co/ca-prov-srv-01-impd-co.db
  2500000000 100%  143.33MB/s    0:00:16 (xfer#1, to-check=3/9)
ca-prov-srv-01-impd-inc/
ca-prov-srv-01-impd-inc/ca-prov-srv-01-impd-inc.db
  2500000000 100%  153.50MB/s    0:00:15 (xfer#2, to-check=2/9)
ca-prov-srv-01-impd-main/
ca-prov-srv-01-impd-main/ca-prov-srv-01-impd-main.db
  2500000000 100%  132.17MB/s    0:00:18 (xfer#3, to-check=1/9)
ca-prov-srv-01-impd-notify/
ca-prov-srv-01-impd-notify/ca-prov-srv-01-impd-notify.db
  2500000000 100%  130.91MB/s    0:00:18 (xfer#4, to-check=0/9)

sent 137 bytes  received 9810722 bytes  139161.12 bytes/sec
total size is 10000000000  speedup is 1019.28
27.237u 5.696s 1:09.43 47.4%    0+0k 128+19531264io 2pf+0w
[dsa@pwdha03 ~/data]$ ls
ca-prov-srv-01-impd-co  ca-prov-srv-01-impd-inc  ca-prov-srv-01-impd-main  ca-prov-srv-01-impd-notify
[dsa@pwdha03 ~/data]$ find $DXHOME/data/ -mindepth 1 -type d -exec bash -c 'mv  $0 ${0/01/03}' {} \; > & /dev/null
[dsa@pwdha03 ~/data]$ ls
ca-prov-srv-03-impd-co  ca-prov-srv-03-impd-inc  ca-prov-srv-03-impd-main  ca-prov-srv-03-impd-notify
[dsa@pwdha03 ~/data]$ find $DXHOME/data -depth -name '*.db' -exec bash -c 'mv  $0 ${0/01/03}' {} \; > & /dev/null
[dsa@pwdha03 ~/data]$ dxserver start all
Starting all dxservers
ca-prov-srv-03-impd-main starting
..
ca-prov-srv-03-impd-main started
ca-prov-srv-03-impd-notify starting
..
ca-prov-srv-03-impd-notify started
ca-prov-srv-03-impd-co starting
..
ca-prov-srv-03-impd-co started
ca-prov-srv-03-impd-inc starting
..
ca-prov-srv-03-impd-inc started
ca-prov-srv-03-imps-router starting
..
ca-prov-srv-03-imps-router started
[dsa@pwdha03 ~/data]$ du -hs
9.4G    .
[dsa@pwdha03 ~/data]$


Note: An enhancement has been open to request that the ‘dsa’ userID is able to use remote SSH processes to address any challenges if the Data IMPD DSAs need to be copied or retained for backup processes.

https://community.broadcom.com/participate/ideation-home/viewidea?IdeationKey=7c795c51-d028-4db8-adb1-c9df2dc48bff

Example for vApp Patches:

Note: There is no major different in speed if the files being copied are already compressed. The below image shows that initial copy is at the rate of the network w/ latency. The value gain from using rsync is still the checksum feature that allow auto-restart where it left off.

vApp Patch process refined to a few lines (to three nodes of a cluster deployment)

# PATCHES
# On Local vApp [as config userID]
mkdir -p patches  && cd patches
curl -L -O ftp://ftp.ca.com/pub/CAIdentitySuiteVA/cumulative-patches/14.3.0/CP-VA-140300-0002.tar.gpg
curl -L -O ftp://ftp.ca.com/pub/CAIdentitySuiteVA/cumulative-patches/14.3.0/CP-IMV-140300-0001.tgz.gpg
screen    [will open a new bash shell ]
patch_vapp CP-VA-140300-0002.tar.gpg           [Patch VA prior to any solution patch]
patch_vapp CP-IMV-140300-0001.tgz.gpg
exit          [exit screen]
cd ..
# Push from one host to another via scp
IP=192.168.242.136;scp -r patches  config@$IP:
IP=192.168.242.137;scp -r patches  config@$IP:
# Push from one host to another via rsync over ssh          [Minor gain for compressed files]
IP=192.168.242.136;rsync --progress -e 'ssh -ax' -avz $HOME/patches config@$IP:
IP=192.168.242.137;rsync --progress -e 'ssh -ax' -avz $HOME/patches config@$IP:
# Pull from one host to another via rsync over ssh          [Minor gain for compressed files]
IP=192.168.242.135;rsync --progress -e 'ssh -ax' -avz config@$IP:./patches $HOME

# View the files were patched
IP=192.168.242.136;ssh -tt config@$IP "ls -lart patches"
IP=192.168.242.137;ssh -tt config@$IP "ls -lart patches"

# On Remote vApp Node #2
IP=192.168.242.136;ssh $IP
cd patches
screen    [will open a new bash shell ]
patch_vapp CP-VA-140300-0002.tar.gpg
patch_vapp CP-IMV-140300-0001.tgz.gpg
exit          [exit screen]
exit          [exit to original host]

# On Remote vApp Node #3
IP=192.168.242.137;ssh $IP
cd patches
screen    [will open a new bash shell ]
patch_vapp CP-VA-140300-0002.tar.gpg
patch_vapp CP-IMV-140300-0001.tgz.gpg
exit          [exit screen]
exit          [exit to original host]

View of rotating the SSH RSA key for CONFIG User ID

# CONFIG - On local vApp host
ls -lart .ssh     [view any prior files]
echo y | ssh-keygen -b 4096 -N Password01 -C $USER -f $HOME/.ssh/id_rsa
IP=192.168.242.135;ssh-keyscan -p 22 $IP >> .ssh/known_hosts
IP=192.168.242.136;ssh-keyscan -p 22 $IP >> .ssh/known_hosts
IP=192.168.242.137;ssh-keyscan -p 22 $IP >> .ssh/known_hosts
cp -r -p .ssh/id_rsa.pub .ssh/authorized_keys
rm -rf /tmp/*.$USER.ssh-keys.tar
tar -cvf /tmp/`/bin/date -u +%s`.$USER.ssh-keys.tar .ssh
ls -lart /tmp/*.$USER.ssh-keys.tar
eval `ssh-agent` && ssh-add           [Enter Password for SSH RSA Private Key]
IP=192.168.242.136;scp `ls /tmp/*.$USER.ssh-keys.tar`  config@$IP:
IP=192.168.242.137;scp `ls /tmp/*.$USER.ssh-keys.tar`  config@$IP:
USER=config;ssh -tt $USER@192.168.242.136 "tar -xvf *.$USER.ssh-keys.tar"
USER=config;ssh -tt $USER@192.168.242.137 "tar -xvf *.$USER.ssh-keys.tar"
IP=192.168.242.136;ssh $IP `/bin/date -u +%s`
IP=192.168.242.137;ssh $IP `/bin/date -u +%s`
IP=192.168.242.136;ssh -vv $IP              [Use -vv to troubleshoot ssh process]
IP=192.168.242.137;ssh -vv $IP 				[Use -vv to troubleshoot ssh process]

Be safe and automate your backups for CA Directory Data DSAs to LDIF

The CA Directory solution provides a mechanism to automate daily on-line backups, via one simple parameter:

dump dxgrid-db period 0 86400;

Where the first number is the offset from GMT/UTC (in seconds) and the second number is how often to run the backup (in seconds), e.g. Once a day = 86400 sec = 24 hr x 60 min/hr x 60 sec/min

Two Gaps/Challenge(s):

History: The automated backup process will overwrite the existing offline file(s) (*.zdb) for the Data DSA. Any requirement or need to perform a RCA is lost due to this fact. What was the data like 10 days ago? With the current state process, only the CA Directory or IM logs would be of assistance.

Size: The automated backup will create an offline file (*.zdb) footprint of the same size as the data (*.db) file. If your Data DSA (*.db) is 10 GB, then your offline (*.zdb) will be 10 GB. The Identity Provisioning User store has four (4) Data DSAs, that would multiple this number , e.g. four (4) db files + four (4) offline zdb files at 10 GB each, will require minimal of 80 GB disk space free. If we attempt to retain a history of these files for fourteen (14) days, this would be four (4) db + fourteen (14) zdb = eighteen (18) x 10 GB = 180 GB disk space required.

Resolutions:

Leverage the CA Directory tool (dxdumpdb) to convert from the binary data (*.db/*.zdb) to LDIF and the OS crontab for the ‘dsa’ account to automate a post ‘online backup’ export and conversion process.

Step 1: Validate the ‘dsa’ user ID has access to crontab (to avoid using root for this effort). cat /etc/cron.allow

If access is missing, append the ‘dsa’ user ID to this file.

Step 2: Validate that online backup process have been scheduled for your Data DSA. Use a find command to identify the offline files (*.zdb ). Note the size of the offline Data DSA files (*.zdb).

Step 3: Identify the online backup process start time, as defined in the Data DSA settings DXC file or perhaps DXI file. Convert this GMT offset time to the local time on the CA Directory server. (See references to assist)

Step 4: Use crontab -e as ‘dsa’ user ID, to create a new entry: (may use crontab -l to view any entries). Use the dxdumpdb -z switch with the DSA_NAME to create the exported LDIF file. Redirect this output to gzip to automatically bypass any need for temporary files. Note: Crontab has limited variable expansion, and any % characters must be escaped.

Example of the crontab for ‘dsa’ to run 30 minutes after (at 2 am CST) the online backup process is scheduled (at 1:30 am CST).

# Goal:  Export and compress the daily DSA offline backup to ldif.gz at 2 AM every day
# - Ensure this crontab runs AFTER the daily automated backup (zdb) of the CA Directory Data DSAs
# - Review these two (2) tokens for DATA DSAs:  ($DXHOME/config/settings/impd.dxc  or ./impd_backup.dxc)
#   a)   Location:  set dxgrid-backup-location = "/opt/CA/Directory/dxserver/backup/";
#   b)   Online Backup Period:   dump dxgrid-db period 0 86400;
#
# Note1: The 'N' start time of the 'dump dxgrid-db period N M' is the offset in seconds from midnight of UTC
#   For 24 hr clock, 0130 (AM) CST calculate the following in UTC/GMT =>  0130 CST + 6 hours = 0730 UTC
#   Due to the six (6) hour difference between CST and UTC TZ:  7.5 * 3600 = 27000 seconds
# Example(s):
#   dump dxgrid-db period 19800 86400;   [Once a day at 2330 CST]
#   dump dxgrid-db period 27000 86400;   [Once a day at 0130 CST]
#
# Note2:  Alternatively, may force an online backup using this line:
#               dump dxgrid-db;
#        & issuing this command:  dxserver init all
#
#####################################################################
#        1      2         3       4       5        6
#       min     hr      d-o-m   month   d-o-w   command(s)
#####################################################################
#####
#####  Testing Backup Every Five (5) Minutes ####
#*/5 * * * *  . $HOME/.profile && dxdumpdb -z `dxserver status | grep "impd-main" | awk "{print $1}"` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-main" | awk '{print $1}'`_`/bin/date --utc +\%Y\%m\%d\%H\%M\%S.0Z`.ldif.gz
#####
#####  Backup daily at 2 AM CST  -  30 minutes after the online backup at 1:30 AM CST #####
#####
0 2 * * *    . $HOME/.profile &&  dxdumpdb -z `dxserver status | grep "impd-main"   | awk "{print $1}"` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-main"   | awk '{print $1}'`_`/bin/date --utc +\%Y\%m\%d\%H\%M\%S.0Z`.ldif.gz
0 2 * * *    . $HOME/.profile &&  dxdumpdb -z `dxserver status | grep "impd-co"     | awk "{print $1}"` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-co"     | awk '{print $1}'`_`/bin/date --utc +\%Y\%m\%d\%H\%M\%S.0Z`.ldif.gz
0 2 * * *    . $HOME/.profile &&  dxdumpdb -z `dxserver status | grep "impd-inc"    | awk "{print $1}"` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-inc"    | awk '{print $1}'`_`/bin/date --utc +\%Y\%m\%d\%H\%M\%S.0Z`.ldif.gz
0 2 * * *    . $HOME/.profile &&  dxdumpdb -z `dxserver status | grep "impd-notify" | awk "{print $1}"` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-notify" | awk '{print $1}'`_`/bin/date --utc +\%Y\%m\%d\%H\%M\%S.0Z`.ldif.gz

Example of the above lines that can be placed in a bash shell, instead of called directly via crontab. Note: Able to use variables and no need to escape the `date % characters `

# set DSA=main &&   dxdumpdb -z `dxserver status | grep "impd-$DSA" | awk '{print $1}'` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-$DSA" | awk '{print $1}'`_`/bin/date --utc +%Y%m%d%H%M%S.0Z`.ldif.gz
# set DSA=co &&     dxdumpdb -z `dxserver status | grep "impd-$DSA" | awk '{print $1}'` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-$DSA" | awk '{print $1}'`_`/bin/date --utc +%Y%m%d%H%M%S.0Z`.ldif.gz
# set DSA=inc &&    dxdumpdb -z `dxserver status | grep "impd-$DSA" | awk '{print $1}'` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-$DSA" | awk '{print $1}'`_`/bin/date --utc +%Y%m%d%H%M%S.0Z`.ldif.gz
# set DSA=notify && dxdumpdb -z `dxserver status | grep "impd-$DSA" | awk '{print $1}'` | gzip -9 > /tmp/`hostname`_`dxserver status | grep "impd-$DSA" | awk '{print $1}'`_`/bin/date --utc +%Y%m%d%H%M%S.0Z`.ldif.gz
#

Example of the output:

Monitor with tail -f /var/log/cron (or syslog depending on your OS version), when the crontab is executed for your ‘dsa’ account

View the output folder for the newly created gzip LDIF files. The files may be extracted back to LDIF format, via gzip -d file.ldif.gz. Compare these file sizes with the original (*.zdb) files of 2GB.

Recommendation(s):

Implement a similar process and retain this data for fourteen (14) days, to assist with any RCA or similar analysis that may be needed for historical data. Avoid copied the (*.db or *.zdb) files for backup, unless using this process to force a clean sync between peer MW Data DSAs.

The Data DSAs may be reloaded (dxloadb) from these LDIF snapshots; the LDIF files do not have the same file size impact as the binary db files; and as LDIF files, they may be quickly search for prior data using standard tools such as grep “text string” filename.ldif.

This process will assist in site preparation for a DAR (disaster and recovery) scenario. Protect your data.

References:

dxdumpdb

https://techdocs.broadcom.com/content/broadcom/techdocs/us/en/ca-enterprise-software/layer7-identity-and-access-management/directory/14-1/administrating/tools-to-manage-ca-directory/dxtools/dxdumpdb-tool-export-data-from-a-datastore-to-an-ldif-file.html

dump dxgrid-db

https://techdocs.broadcom.com/content/broadcom/techdocs/us/en/ca-enterprise-software/layer7-identity-and-access-management/directory/14-1/reference/commands-reference/dump-dxgrid-db-command-take-a-consistent-snapshot-copy-of-a-datastore.html

If you wish to learn more or need assistance, contact us.