Skip to main content

Release Notes

Latest Agent Version: Server v2.9.27​


2026-11-05

v2.9.27

New Added support for regex capture values ($1, $2, etc.) in process group names.

  • A single rule can create a separate group for each capture value (e.g., one group per executable JAR name).

New Added per-file regex filters for log collection (logsink) targets.

New Added agent activation information for the transition to Category V2.

Fixed Improved the infra_process_group and infra_process_topn categories.

  • Fixed cpu and diskIo values intermittently dropping to 0 or doubling on Linux.
  • Added the memory usage field (pWorkingSet, %) on Windows.
  • Added WhaTap project information linked to each process (link*) to the infra_process_group category on Linux.

Fixed Fixed the agent terminating abnormally.

  • Fixed the agent terminating due to concurrent access while monitoring event logs on Windows.
  • Fixed the agent terminating when netstat fails to run on Windows.
  • Fixed the agent terminating when Veritas InfoScale vxstat output contains blank lines on Linux.
  • Fixed the watchdog terminating when watchdog connection errors repeat on Linux.

Fixed Fixed handles not being released when querying event logs on Windows.

Fixed Improved processing of custom category script results.

  • Whitespace around delimiters is now trimmed so that numeric values and headers are recognized correctly.

Fixed Added the GPU count field (gpu_count) to the infra_gpu category.

Fixed Fixed a log message about missing VM collection targets being written on servers that do not use vCenter/ESXi.

2026-10-21

v2.9.26

New Added APM and DPM connection information for each process to the infra_process_topn category on Linux.

Fixed Applied fixes from static code analysis.

  • Fixed the agent terminating abnormally.

Fixed Fixed collection stopping when integrating with a large vCenter.

  • Fixed the collection timeout being applied as short as the connection timeout.

Fixed Fixed pRss in the infra_process_group category being incorrectly calculated as an average on Linux.

Fixed Reduced CPU usage of log monitoring.

Fixed Added the usage of the most-used file system (pUsedMax) and its mount path (mountName) to the infra_disk category.

Fixed Fixed agent installation failing on Debian 13.

  • Fixed the repository signing key being rejected by recent OS versions because it used SHA-1.

2026-09-30

v2.9.25

New Added remote agent stop and restart from OPM.

Fixed Improved the server_system_topology category.

  • Added CPU type, physical core count, logical core count, and memory capacity metrics.
  • Added GPU index, model name, UUID, and RDMA affinity information.
  • Added link layer, rate, and link/phys state information for RDMA NICs.
  • Sends the physicalCores and pcieDepth fields as numeric types.

Fixed Improved the server_gpu_topology category.

  • Added GPU BDF (baseGpuBdf, targetGpuBdf), UUID, and pcieGpuCount information.
  • Aligned the GPU BDF format with pciBusId in server_system_topology.
  • Reports QUERY_FAILED when a P2P query fails, so it is distinct from the NVML Unknown (U) state.

Fixed Improved VM-related metrics.

  • Fixed the vHost performance category not being collected on Windows.
  • Fixed VM collection settings, such as credentials, not applying immediately on Windows.
  • Fixed removal and addition history piling up alternately in the inventory every cycle when several vCenters are connected.
  • Fixed the VM metering core count being undercounted when several vCenters are connected. A vHost registered in more than one vCenter is now counted once.
  • Sends memory fields of the infra_vm_vhost_vmguest_perf category in bytes, and disk and network usage in bytes per second.
  • Sends the four overcommit fields of the infra_vm_vhost_resource_capacity category as percentages instead of ratios.
  • Added the oname tag to every vHost performance category.

Fixed Fixed traffic not being collected for NICs assigned only an IPv6 address.

Fixed Improved unnecessary agent restarts caused by transient Windows PDH errors.

  • A transient error now skips only that collection cycle, and collection continues.
  • The error state clears automatically once collection returns to normal.
  • Added the pdh.error.restart.limit option. The agent restarts only after PDH metric collection fails this many times in a row (default 12, about one minute at a 5-second interval).

Fixed Improved process user information handling (server_process, infra_process_group).

  • Accounts missing from /etc/passwd, such as SSSD and LDAP accounts, are shown by numeric uid.
  • Fixed NIS/LDAP compat lines (+::::::) being matched to the root account.

Fixed Added diagnostic logs for Windows installer security verification failures.

2026-09-04

v2.9.24

Fixed Improved VM-related metrics.

  • Supports configuring VM options from the UI.
  • Optimized the data structure.
  • Supports connecting multiple vCenter and vHost instances from a single agent.
  • Added metrics for the infra_vm_vhost_mem category.

Fixed Improved the whatap.conf synchronization logic.

  • Fixed the file modification time changing because the file was rewritten even without changes.

Fixed Added the Nutanix signature to Windows virtualization detection.

2026-08-19

v2.9.23

New Added an option to exclude mount points from server_disk and infra_filesystem collection.

  • filesystem.exclude.pattern: Specifies the mount points to exclude, using a regular expression.

New Added inspection integration logic.

  • Separated the data transmission paths for inspection and cronjob.
  • Added inspection pcode information to server_inventory.
  • Restricted transmission to custom data only.

Fixed Improved VM metrics.

  • Added metrics to the infra_vm_vhost_nic_perf, infra_vm_vhost_network, infra_vm_vhost_cpu, and infra_vm_vhost_vmguest_perf categories.
  • Added the new infra_vm_vhost_hw and infra_vm_vhost_resource_capacity categories.
  • Added vm_vcenter_inventory inventory information.
  • Added VM metering information.

Fixed Improved the infra_infiniband category.

Fixed Improved the infra_process_topn category.

  • Provides the pod, container, and namespace information associated with each process.

Fixed Improved watchdog process logging.

Fixed Added error handling for duplicate cronjob script names.

Fixed Added strftime format parsing for file names in file monitoring event settings.

Fixed Fixed incorrect permissions being set when creating a cronjob path.

Fixed Improved cronjob row_id handling.

  • Fixed the value being passed as a string.
  • Fixed duplicate hash values when sending data across multiple rows.

Fixed Improved the agent patch feature.

2026-08-05

v2.9.22

New Added an option to adjust the housekeeping agent restart time in minutes.

  • housekeeping.hour: default 2
  • housekeeping.minute: default 0

Fixed Improved categoryv2 metrics.

  • Added new categories.
    • server_rdma_inventory: RDMA configuration information.
    • server_gpu_topology: GPU connection topology information.
  • Improved existing categories.
    • infra_cpu_core: Added NUMA node information.
    • infra_nic_info: Improved Windows teaming and Linux bonding information.
    • infra_numa_inventory: Added GPU-related information.

Fixed Fixed an issue where changing a cronjob setting refreshed the entire interval condition.

Fixed Improved RPC communication between the Windows watchdog and foreground process.

  • Fixed an issue where a named pipe write failure forcibly closed the connection and stopped the agent for 3 minutes.
    • Added retry handling on write failure.
    • Added a fast-restart path when retries also fail.
  • Added an option to change the RPC communication method.
    • housekeeping.rpc.mode: Can specify socket (default: pipe).
    • housekeeping.rpc.port: Specifies the port to use in socket mode (default: 6601).

Fixed Fixed an issue where Mig Used Slice information was incorrect when using the nvidiav2.collector option.

Fixed Fixed excessive NetHeartbeat log output.

2026-07-23

v2.9.21

Fixed Improved categoryv2 metrics.

  • Added new categories.
    • server_numa_inventory: NUMA inventory information.
    • infra_numa_perf: NUMA performance metrics.
    • infra_os_parameter: OS parameters. Use the os.parameter.custom option to collect additional parameters beyond the defaults (comma-separated).
    • infra_container: Container metrics. Requires the container.enabled option (default: false). Covers Docker containers, with further expansion planned.
  • Improved existing categories.
    • infra_filesystem: Now provides NFS filesystem server information and full ZFS dataset paths.
    • infra_ntp_info: Fixed abnormal data on Windows.
    • infra_nic_info: Improved NIC collection on Windows.
    • infra_process_topn: Improved process environment variable collection. Use the process.topn.envkeys option to specify environment variables to collect (up to 5).

Fixed Added a ver tag to folding (recType: 2) data.

  • Prevents unfolding failures caused by field spec changes from category improvements.

Fixed Fixed a CPU core count mismatch in inventory on virtualized environments.

Fixed Fixed an issue where the server_ntpq metric could not be collected in chronyc environments.

Fixed Added a row_id tag when creating a cronjob custom category.

Fixed Fixed an issue where the domain was shown together with Windows process owner information.

Fixed Changed the default value of the ipreg option to false.

  • Stays true when the autoscale option is applied.

Fixed Improved a certificate validation failure case for the Windows Installer in closed network environments.

Fixed Strengthened logging for Windows network errors and agent crash cases.

2026-07-13

v2.9.20

New Added an option to specify the user that runs cronjob.

New Added support for Windows-style cronjob schedule options.

  • Runs a job periodically on a specific day and time — daily, weekly, monthly, or yearly.

New When using the cronjob folding option, the key (@id) is now generated automatically based on the specified column.

Fixed Improved the infra_process_group custom option.

  • Added data for event processing.
  • Explicitly returns data (count = 0) when no matching information exists.

Fixed Changed the pk0 generation rule for cronjob results saved to kvTable.

  • Improved to be based on the specified fields, consistent with tagCount.

Fixed Changed the pk0 type in agent_inventory.

  • int64 → int32

Fixed Improved a Windows installer vulnerability.

Fixed Added pk0 information for the target GPU in infra_gpu_xid.

Fixed Improved settings sync for file keyword, Windows event, and file watch alert configurations.

  • If settings changed while the agent was stopped, they now sync automatically on startup.

Fixed Added hostname and osType information to settings sync requests.

Fixed Fixed an issue where container data could not be collected after the Docker root directory changed.

  • Added dynamic path tracking.
  • Added the docker.root.dir option in whatap.conf to specify the root directory.

Fixed Fixed an issue where some metrics were not collected on Windows when housekeeping.enabled=false was applied.

Fixed Fixed an issue where the infra_win_service category could not be collected after code obfuscation.

Fixed Added a collection threshold option for PSS and USS.

  • Identified cases where collecting data from high-memory processes (e.g. in-memory databases) delayed the service.
  • Only collects items with RSS of 4GB or less.
  • Adjustable via the proc.pss.threshold option (in bytes).

Fixed Changed the collection unit for readByte/writeByte in infra_zpool_perf.

  • KB/s → Byte/s, for consistency with other metrics.

2026-06-15

v2.9.19

Fixed Fixed an issue where running the whatap-updater service status command on RHEL6-based systems logged errors to /var/log/messages

Fixed Improved CPU usage during Process data collection

Fixed Fixed disk collection error case

Fixed Added infra_process_group custom field

  • Field separator for entries created via the procgroup option

2026-06-08

v2.9.18

New Improved server inventory fields

  • Provides uptime in daily units.
  • Fixed incorrect GPU count display.

New Improved GPU inventory fields

  • Added 1-hour cumulative power field.

New Added GPU XID error information collection

Fixed procgroup option bug fix

  • Fixed deleted rules remaining with count 0.
  • Fixed missing name field when grouping with kernel-processgroup in the agent.

Fixed Fixed ZFS IO collection failure cases

Fixed Improved OPM script single execution failure cases

  • Fixed cases where long scripts could not be passed as shell parameters.
  • Local scripts can now be executed by filename.

Fixed Fixed OPM patch request failure when installed with --prefix option using tar.gz

2026-05-21

v2.9.17

New Added custom option support for infra_process_group.

  • Option format: procgrop_[gropuid]=[groupName]@@[user]@@[pattern]@@[except pattern]
  • UI support is planned.

Fixed Improved infra_gpu_process metrics.

  • Added Physical GPU link information when MIG is enabled.
  • Fixed cases where the totalCharIo field was filled with 0.
  • Improved @id key generation rules.

Fixed Fixed cases where data collection failed when setting the Windows event log level.

Fixed Fixed cases where external IP was collected as local IP in Windows Inventory.

Fixed Changed infra_gpu power usage collection method from average to sum.

2026-05-12

v2.9.16

Fixed Added MIG Instance support for GPU DRAM Active metrics.

Fixed Changed Process cmd2 length constraint options.

  • server.proc.cmd2.max: Constrains server_process data, default unlimit
  • proc.cmd2.max: Constrains infra_gpu_topn data, default 256

2026-05-11

v2.9.15

New Improved GPU metrics collection

  • Uses the improved metric collection method when nvidiav2.collector=true is set.
  • Switched from command-based collection (nvidia-smi, dcgmi) to library-based collection (NVML, DCGM).
  • Added derived metrics such as violations.
  • Provides GPU process information at the MIG level.

New Added Ethernet type support for the infra_infiniband category.

New Added new categoryv2 categories and metrics

  • infra_vxlv_perf (enable with vxlv.perf.enabled=true).
  • infra_zpool_perf (enable with zpool.perf.enabled=true).

Fixed Fixed Hangul corruption when watching Linux EUC-KR file keywords.

Fixed Fixed excessive memory usage on hosts with many CPU cores

  • Limits memory usage by applying GOMAXPROCS and MALLOC_ARENA_MAX.

Fixed Added option to disable disk IO collection

  • Resolves resource overhead during IO collection on hosts with many disks.
  • Setting disk.io.enabled=false excludes per-disk IO metrics from collection.

Fixed Fixed service error output when restarting the agent on RHEL.

2026-04-29

v2.9.14

New Added a tar.gz package to support custom installation paths on Ubuntu.

  • The deb package does not support installation path customization.
  • A separate tar.gz package is provided.

New Added infrastructure for WhaTap Agent Meta information collection.

  • When each agent provides Meta integration data, the related information is automatically collected.

New Added new categories and metrics for categoryv2.

New Reduced CPU load when a large number of disks are present.

  • Optimized resource usage.
  • Added disk.mount.whitelist and disk.mount.blacklist options.
  • Collects only required mount points by filtering.
  • Use a comma (,) as the delimiter when specifying multiple values.

Fixed Fixed service-related error output cases on RHEL7.

Fixed Fixed false positive cases in file keyword monitoring.

Fixed Fixed increased agent memory usage after applying the code obfuscation option.

Fixed Added support for multi-character, Unicode, and tab characters as Cronjob delimiters.

2026-03-31

v2.9.13

Fixed Improved security vulnerability issues

  • Updated default Go version
  • Updated vulnerable library versions
  • Removed debugging symbols and added obfuscation
  • Added Anti-debug and file integrity verification to Windows Installer

Fixed Added inventory information delivery when using the Inspection license option

2026-03-09

v2.9.12

New Added Process PSS and USS metrics

  • OS version constraints (RHEL 8 or later, Ubuntu 18.04 or later, Debian 10 or later)

New Added infra_process category data

  • Process status count and File Descriptor related information

New Added license option for preventive maintenance

Fixed Fixed abnormal behavior of agent patch function when using the prefix option during RPM installation

2026-03-03

v2.9.11

New Added option to disable agent data collection

  • Temporary measure for collection server overload situations

Fixed Improved gpu_inventory data

  • Measures for front-end performance improvement

Fixed Fixed Windows patch failure cases

Fixed Improved Network Error cases during data collection overload

  • Asynchronous processing for cases where data collection may take long depending on the environment such as process, disk, and network

Fixed Fixed core information mismatch on servers with 256 or more cores

Fixed Fixed infra_process_topn data overload cases

  • Added cmd2 length limit (default 256)

2026-02-10

v2.9.10

Fixed Fixed cases where Windows log events were not generated

Fixed Fixed Windows patch failure cases

Fixed Fixed Path Traversal vulnerability issue

Fixed Fixed installation failure on Debian 7 due to GLIBC constraints

Fixed Fixed GPU temperature not being collected when DCGM service is not in use

2026-01-28

v2.9.9

Fixed Improved GPU inventory data transmission structure

  • Applied inventory data modification/deletion cases

Fixed Removed rpm -qa use case

Fixed Fixed an issue where CentOS6 agent automatic update was not possible

2026-01-20

v2.9.8

New CronJob TagCount Fold option support

New Added infra_mem scan, free, and steal related metrics

Fixed Improved GPU inventory based on NVML Library

  • Disabled by default for backward compatibility, and will be enabled by default after stabilization

Fixed Improved GPU inventory data collection interval

  • Changed from 1 day → 10 minutes (optional)

Fixed Fixed an issue where OS was not displayed in CentOS6 inventory information

2026-01-07

v2.9.7

Agent

New Added Agent Patch feature (Operation Manager)

New Added simplified GPU model information

Fixed Fixed rpmdb corruption issue during agent startup

Fixed Fixed cases where inventory information was displayed incorrectly

2025-11-26

v2.9.6

New Added GPU Utilization metric

Fixed Modified to adjust call frequency when GPU dxgm-exporter hangs

Fixed Fixed issue where metrics were not collected with GPU Driver R440

Fixed Improved timeout issue when collecting Inventory information in GPU MIG environment

Fixed Fixed issue where Physical GPU metrics were not collected when using dcgm-exporter in GPU MIG environment

Last updated on