_R_R_D_C_R_E_A_T_E(1)                         rrdtool                        _R_R_D_C_R_E_A_T_E(1)

NNAAMMEE
     rrdcreate - Set up a new Round Robin Database

SSYYNNOOPPSSIISS
     rrrrddttooooll  ccrreeaattee  _f_i_l_e_n_a_m_e  [----ssttaarrtt|--bb _s_t_a_r_t _t_i_m_e] [----sstteepp|--ss _s_t_e_p] [----tteemm‐‐
     ppllaattee|--tt _t_e_m_p_l_a_t_e_-_f_i_l_e]    [----ssoouurrccee|--rr _s_o_u_r_c_e_-_f_i_l_e]    [----nnoo--oovveerrwwrriittee|--OO]
     [----ddaaeemmoonn|--dd _a_d_d_r_e_s_s]               [DDSS::_d_s_-_n_a_m_e[==_m_a_p_p_e_d_-_d_s_-_n_a_m_e[[[_s_o_u_r_c_e_-_i_n_‐
     _d_e_x]]]]::_D_S_T::_d_s_t _a_r_g_u_m_e_n_t_s] [RRRRAA::_C_F::_c_f _a_r_g_u_m_e_n_t_s]

DDEESSCCRRIIPPTTIIOONN
     The create function of RRDtool lets you set up  new  Round  Robin  Database
     (RRRRDD)  files.   The file is created at its final, full size and filled with
     _*_U_N_K_N_O_W_N_* data, unless one or more source RRRRDD files have been specified and
     they hold suitable data to "pre-fill" the new RRRRDD file.

   _ff_ii_ll_ee_nn_aa_mm_ee
     The name of the RRRRDD you want to create. RRRRDD files should end with  the  ex‐
     tension _._r_r_d. However, RRRRDDttooooll will accept any filename.

   ----ssttaarrtt||--bb _ss_tt_aa_rr_tt _tt_ii_mm_ee ((ddeeffaauulltt:: nnooww -- 1100ss))
     Specifies  the  time  in  seconds since 1970-01-01 UTC when the first value
     should be added to the RRRRDD. RRRRDDttooooll will not accept any data  timed  before
     or at the time specified.

     See  also "AT-STYLE TIME SPECIFICATION" in rrdfetch for other ways to spec‐
     ify time.

     If one or more source files is used to pre-fill the new  RRRRDD,  the  ----ssttaarrtt
     option  may  be  omitted.  In  that  case, the latest update time among all
     source files will be used as the last update time of the new RRRRDD file,  ef‐
     fectively setting the start time.

   ----sstteepp||--ss _ss_tt_ee_pp ((ddeeffaauulltt:: 330000 sseeccoonnddss))
     Specifies the base interval in seconds with which data will be fed into the
     RRRRDD.   A  scaling  factor  may  be  present as a suffix to the integer; see
     "STEP, HEARTBEAT, and Rows As Durations".

   ----nnoo--oovveerrwwrriittee||--OO
     Do not clobber an existing file of the same name.

   ----ddaaeemmoonn||--dd _aa_dd_dd_rr_ee_ss_ss
     Address of the rrdcached daemon.  For a list of accepted formats,  see  the
     --ll option in the rrdcached manual.

      rrdtool create --daemon unix:/var/run/rrdcached.sock /var/lib/rrd/foo.rrd I<other options>

   [[----tteemmppllaattee||--tt _tt_ee_mm_pp_ll_aa_tt_ee_--_ff_ii_ll_ee]]
     Specifies  a  template  RRRRDD file to take step, DS and RRA definitions from.
     This allows one to base the structure of a new file on some existing  file.
     The data of the template file is NOT used for pre-filling, but it is possi‐
     ble to specify the same file as a source file (see below).

     Additional DS and RRA definitions are permitted, and will be added to those
     taken from the template.

   ----ssoouurrccee||--rr _ss_oo_uu_rr_cc_ee_--_ff_ii_ll_ee
     One  or  more  source RRRRDD files may be named on the command line. Data from
     these source files will be used to prefill the created RRRRDD file. The output
     file and one source file may refer to the same file name. This will  effec‐
     tively  replace  the  source file with the new RRRRDD file. While there is the
     danger to loose the source file because it gets replaced, there is no  dan‐
     ger that the source and the new file may be "garbled" together at any point
     in  time,  because  the new file will always be created as a temporary file
     first and will only be moved to its final  destination  once  it  has  been
     written in its entirety.

     Prefilling  is  done  by matching up DS names, RRAs and consolidation func‐
     tions and choosing the best available data resolution when doing  so.  Pre‐
     filling may not be mathematically correct in all cases (e.g. if resolutions
     have  to  change  due to changed stepping of the target RRD and old and new
     resolutions do not match up with old/new bin boundaries in RRAs).

     In other words: A best effort is made to preserve data  during  prefilling.
     Also,  pre-filling  of  RRAs  may  only be possible for certain kinds of DS
     types. Prefilling may also have strange effects on Holt-Winters forecasting
     RRAs. In other words: there is no guarantee for data-correctness.

     When "pre-filling" a RRRRDD file, the structure of the new file must be speci‐
     fied as usual using DS and RRA specifications as outlined below. Data  will
     be taken from source files based on DS names and types and in the order the
     source files are specified in. Data sources with the same name from differ‐
     ent source files will be combined to form a new data source. Generally, for
     any point in time the new RRRRDD file will cover after its creation, data from
     only  one  source  file  will have been used for pre-filling. However, data
     from multiple sources may be combined if it refers to different times or an
     earlier named source file holds unknown data for a time where a  later  one
     holds known data.

     If  this  automatic data selection is not desired, the DS syntax allows one
     to specify a mapping of target and source data sources for prefilling. This
     syntax allows one to rename data sources and to restrict prefilling  for  a
     DS to only use data from a single source file.

     Prefilling  currently only works reliably for RRAs using one of the classic
     consolidation functions, that is one of: AVERAGE, MIN, MAX, LAST. It  might
     also currently have problems with COMPUTE data sources.

     Note  that  the  act of prefilling during ccrreeaattee is similar to a lot of the
     operations available via the ttuunnee command, but using ccrreeaattee syntax.

   DDSS::_dd_ss_--_nn_aa_mm_ee[[==_mm_aa_pp_pp_ee_dd_--_dd_ss_--_nn_aa_mm_ee[[[[_ss_oo_uu_rr_cc_ee_--_ii_nn_dd_ee_xx]]]]]]::_DD_SS_TT::_dd_ss_tt _aa_rr_gg_uu_mm_ee_nn_tt_ss
     A single RRRRDD can accept input from several data sources (DDSS),  for  example
     incoming and outgoing traffic on a specific communication line. With the DDSS
     configuration  option  you  must  define some basic properties of each data
     source you want to store in the RRRRDD.

     _d_s_-_n_a_m_e is the name you will use to reference this particular  data  source
     from  an  RRRRDD.  A _d_s_-_n_a_m_e must be 1 to 19 characters long in the characters
     [a-zA-Z0-9_].

     _D_S_T defines the Data Source Type. The remaining arguments of a data  source
     entry depend on the data source type. For GAUGE, COUNTER, DERIVE, DCOUNTER,
     DDERIVE and ABSOLUTE the format for a data source entry is:

     DDSS::_d_s_-_n_a_m_e::{_G_A_U_G_E   _|   _C_O_U_N_T_E_R  _|  _D_E_R_I_V_E  _|  _D_C_O_U_N_T_E_R  _|  _D_D_E_R_I_V_E  _|  _A_B_‐
     _S_O_L_U_T_E}::_h_e_a_r_t_b_e_a_t::_m_i_n::_m_a_x

     For COMPUTE data sources, the format is:

     DDSS::_d_s_-_n_a_m_e::_C_O_M_P_U_T_E::_r_p_n_-_e_x_p_r_e_s_s_i_o_n

     In order to decide which data source type to use,  review  the  definitions
     that  follow.  Also consult the section on "HOW TO MEASURE" for further in‐
     sight.

     GGAAUUGGEE
         is for things like temperatures or number of people in a  room  or  the
         value of a RedHat share.

     CCOOUUNNTTEERR
         is  for continuous incrementing counters like the ifInOctets counter in
         a router. The CCOOUUNNTTEERR data source assumes that the  counter  never  de‐
         creases,  except  when  a counter overflows.  The update function takes
         the overflow into account.  The counter is stored as a per-second rate.
         When the counter overflows, RRDtool checks if the overflow happened  at
         the 32bit or 64bit border and acts accordingly by adding an appropriate
         value to the result.

     DDCCOOUUNNTTEERR
         the  same  as CCOOUUNNTTEERR, but for quantities expressed as double-precision
         floating point number.  Could be used to track quantities  that  incre‐
         ment  by  non-integer numbers, i.e. number of seconds that some routine
         has taken to run, total weight processed by some  technology  equipment
         etc.   The  only  substantial difference is that DDCCOOUUNNTTEERR can either be
         upward counting or downward counting, but not both at  the  same  time.
         The current direction is detected automatically on the second non-unde‐
         fined counter update and any further change in the direction is consid‐
         ered  a  reset.   The  new direction is determined and locked in by the
         second update after reset and its difference to the value at reset.

     DDEERRIIVVEE
         will store the derivative of the line going from the last to  the  cur‐
         rent value of the data source. This can be useful for gauges, for exam‐
         ple,  to  measure the rate of people entering or leaving a room. Inter‐
         nally, derive works exactly like COUNTER but without  overflow  checks.
         So if your counter does not reset at 32 or 64 bit you might want to use
         DERIVE and combine it with a MIN value of 0.

     DDDDEERRIIVVEE
         the  same  as  DDEERRIIVVEE, but for quantities expressed as double-precision
         floating point number.

         NNOOTTEE oonn CCOOUUNNTTEERR vvss DDEERRIIVVEE

         by Don Baarda <don.baarda@baesystems.com>

         If you cannot tolerate ever mistaking the occasional counter reset  for
         a  legitimate counter wrap, and would prefer "Unknowns" for all legiti‐
         mate counter wraps and resets, always use DERIVE with min=0. Otherwise,
         using COUNTER with a suitable max will return correct  values  for  all
         legitimate  counter  wraps,  mark some counter resets as "Unknown", but
         can mistake some counter resets for a legitimate counter wrap.

         For a 5 minute step and 32-bit counter, the probability of mistaking  a
         counter reset for a legitimate wrap is arguably about 0.8% per 1Mbps of
         maximum  bandwidth.  Note  that  this equates to 80% for 100Mbps inter‐
         faces, so for high bandwidth interfaces and  a  32bit  counter,  DERIVE
         with  min=0  is  probably preferable. If you are using a 64bit counter,
         just about any max setting will eliminate the possibility of  mistaking
         a reset for a counter wrap.

     AABBSSOOLLUUTTEE
         is  for  counters  which  get reset upon reading. This is used for fast
         counters which tend to overflow. So instead of  reading  them  normally
         you  reset  them  after every read to make sure you have a maximum time
         available before the next overflow. Another usage  is  for  things  you
         count like number of messages since the last update.

     CCOOMMPPUUTTEE
         is for storing the result of a formula applied to other data sources in
         the RRRRDD. This data source is not supplied a value on update, but rather
         its  Primary  Data Points (PDPs) are computed from the PDPs of the data
         sources according to the rpn-expression that defines the formula.  Con‐
         solidation  functions are then applied normally to the PDPs of the COM‐
         PUTE data source (that is the rpn-expression is only applied to  gener‐
         ate  PDPs).  In  database  software,  such data sets are referred to as
         "virtual" or "computed" columns.

     _h_e_a_r_t_b_e_a_t defines the maximum number of seconds that may pass  between  two
     updates  of this data source before the value of the data source is assumed
     to be _*_U_N_K_N_O_W_N_*.

     _m_i_n and _m_a_x define the expected range values for data supplied  by  a  data
     source. If _m_i_n and/or _m_a_x are specified any value outside the defined range
     will  be  regarded  as  _*_U_N_K_N_O_W_N_*. If you do not know or care about min and
     max, set them to U for unknown. Note that min and max always refer  to  the
     processed values of the DS. For a traffic-CCOOUUNNTTEERR type DS this would be the
     maximum and minimum data-rate expected from the device.

     _I_f  _i_n_f_o_r_m_a_t_i_o_n _o_n _m_i_n_i_m_a_l_/_m_a_x_i_m_a_l _e_x_p_e_c_t_e_d _v_a_l_u_e_s _i_s _a_v_a_i_l_a_b_l_e_, _a_l_w_a_y_s _s_e_t
     _t_h_e _m_i_n _a_n_d_/_o_r _m_a_x _p_r_o_p_e_r_t_i_e_s_. _T_h_i_s _w_i_l_l _h_e_l_p _R_R_D_t_o_o_l  _i_n  _d_o_i_n_g  _a  _s_i_m_p_l_e
     _s_a_n_i_t_y _c_h_e_c_k _o_n _t_h_e _d_a_t_a _s_u_p_p_l_i_e_d _w_h_e_n _r_u_n_n_i_n_g _u_p_d_a_t_e_.

     _r_p_n_-_e_x_p_r_e_s_s_i_o_n  defines  the  formula used to compute the PDPs of a COMPUTE
     data source from other data sources in the same <RRD>.  It  is  similar  to
     defining a CCDDEEFF argument for the graph command. Please refer to that manual
     page  for  a  list and description of RPN operations supported. For COMPUTE
     data sources, the following RPN operations are not supported: COUNT,  PREV,
     TIME,  and  LTIME. In addition, in defining the RPN expression, the COMPUTE
     data source may only refer to the names of data source listed previously in
     the create command. This is similar to the restriction that CCDDEEFFs must  re‐
     fer only to DDEEFFs and CCDDEEFFs previously defined in the same graph command.

     When  pre-filling  the  new  RRRRDD file using one or more source RRRRDDs, the DS
     specification may hold an optional mapping after the DS  name.  This  takes
     the  form  of an equal sign followed by a mapped-to DS name and an optional
     source index enclosed in square brackets.

     For example, the DS

      DS:a=b[2]:GAUGE:120:0:U

     specifies that the DS named _a should be pre-filled from the DS named  _b  in
     the second listed source file (source indices are 1-based).

   RRRRAA::_CC_FF::_cc_ff _aa_rr_gg_uu_mm_ee_nn_tt_ss
     The  purpose  of an RRRRDD is to store data in the round robin archives (RRRRAA).
     An archive consists of a number of data values or statistics  for  each  of
     the defined data-sources (DDSS) and is defined with an RRRRAA line.

     When  data  is  entered into an RRRRDD, it is first fit into time slots of the
     length defined with the --ss option, thus becoming a _p_r_i_m_a_r_y _d_a_t_a _p_o_i_n_t.

     The data is also processed with the  consolidation  function  (_C_F)  of  the
     archive. There are several consolidation functions that consolidate primary
     data points via an aggregate function: AAVVEERRAAGGEE, MMIINN, MMAAXX, LLAASSTT.

     AVERAGE
         the average of the data points is stored.

     MIN the smallest of the data points is stored.

     MAX the largest of the data points is stored.

     LASTthe last data points is used.

     Note that data aggregation inevitably leads to loss of precision and infor‐
     mation. The trick is to pick the aggregate function such that the _i_n_t_e_r_e_s_t_‐
     _i_n_g properties of your data is kept across the aggregation process.

     The format of RRRRAA line for these consolidation functions is:

     RRRRAA::{_A_V_E_R_A_G_E _| _M_I_N _| _M_A_X _| _L_A_S_T}::_x_f_f::_s_t_e_p_s::_r_o_w_s

     _x_f_f  The xfiles factor defines what part of a consolidation interval may be
     made up from _*_U_N_K_N_O_W_N_* data while the consolidated value is still  regarded
     as  known. It is given as the ratio of allowed _*_U_N_K_N_O_W_N_* PDPs to the number
     of PDPs in the interval. Thus, it ranges from 0 to 1 (exclusive).

     _s_t_e_p_s defines how many of these _p_r_i_m_a_r_y _d_a_t_a _p_o_i_n_t_s are  used  to  build  a
     _c_o_n_s_o_l_i_d_a_t_e_d  _d_a_t_a _p_o_i_n_t which then goes into the archive.  See also "STEP,
     HEARTBEAT, and Rows As Durations".

     _r_o_w_s defines how many generations of data values are kept in an RRRRAA.  Obvi‐
     ously, this has to be greater than zero.  See also  "STEP,  HEARTBEAT,  and
     Rows As Durations".

AAbbeerrrraanntt BBeehhaavviioorr DDeetteeccttiioonn wwiitthh HHoolltt--WWiinntteerrss FFoorreeccaassttiinngg
     In  addition  to  the  aggregate  functions, there are a set of specialized
     functions that enable RRRRDDttooooll to provide data smoothing (via the  Holt-Win‐
     ters  forecasting  algorithm),  confidence bands, and the flagging aberrant
     behavior in the data source time series:

     •   RRRRAA::_H_W_P_R_E_D_I_C_T::_r_o_w_s::_a_l_p_h_a::_b_e_t_a::_s_e_a_s_o_n_a_l _p_e_r_i_o_d[::_r_r_a_-_n_u_m]

     •   RRRRAA::_M_H_W_P_R_E_D_I_C_T::_r_o_w_s::_a_l_p_h_a::_b_e_t_a::_s_e_a_s_o_n_a_l _p_e_r_i_o_d[::_r_r_a_-_n_u_m]

     •   RRRRAA::_S_E_A_S_O_N_A_L::_s_e_a_s_o_n_a_l _p_e_r_i_o_d::_g_a_m_m_a::_r_r_a_-_n_u_m[::ssmmooootthhiinngg--wwiinnddooww==_f_r_a_c_t_i_o_n]

     •   RRRRAA::_D_E_V_S_E_A_S_O_N_A_L::_s_e_a_s_o_n_a_l   _p_e_r_i_o_d::_g_a_m_m_a::_r_r_a_-_n_u_m[::ssmmooootthhiinngg--wwiinnddooww==_f_r_a_c_‐
         _t_i_o_n]

     •   RRRRAA::_D_E_V_P_R_E_D_I_C_T::_r_o_w_s::_r_r_a_-_n_u_m

     •   RRRRAA::_F_A_I_L_U_R_E_S::_r_o_w_s::_t_h_r_e_s_h_o_l_d::_w_i_n_d_o_w _l_e_n_g_t_h::_r_r_a_-_n_u_m

     These  RRRRAAss  differ  from the true consolidation functions in several ways.
     First, each of the RRRRAAs is updated once for every primary data point.  Sec‐
     ond, these  RRRRAAss  are  interdependent.  To  generate  real-time  confidence
     bounds,  a matched set of SEASONAL, DEVSEASONAL, DEVPREDICT, and either HW‐
     PREDICT or MHWPREDICT must exist. Generating smoothed values of the primary
     data points requires a SEASONAL RRRRAA and either an HWPREDICT  or  MHWPREDICT
     RRRRAA.  Aberrant behavior detection requires FAILURES, DEVSEASONAL, SEASONAL,
     and either HWPREDICT or MHWPREDICT.

     The predicted, or smoothed, values are stored in the HWPREDICT  or  MHWPRE‐
     DICT  RRRRAA.  HWPREDICT  and  MHWPREDICT  are  actually two variations on the
     Holt-Winters method. They are interchangeable. Both  attempt  to  decompose
     data  into  three  components:  a baseline, a trend, and a seasonal coeffi‐
     cient.  HWPREDICT adds its seasonal coefficient to the baseline to  form  a
     prediction,  whereas  MHWPREDICT multiplies its seasonal coefficient by the
     baseline to form a prediction. The difference is noticeable when the  base‐
     line  changes  significantly in the course of a season; HWPREDICT will pre‐
     dict the seasonality to stay constant as the baseline changes, but  MHWPRE‐
     DICT  will  predict  the seasonality to grow or shrink in proportion to the
     baseline. The proper choice of method depends on the thing  being  modeled.
     For  simplicity,  the  rest of this discussion will refer to HWPREDICT, but
     MHWPREDICT may be substituted in its place.

     The predicted deviations are stored in DEVPREDICT (think a standard  devia‐
     tion  which  can  be  scaled  to yield a confidence band). The FAILURES RRRRAA
     stores binary indicators. A 1 marks the  indexed  observation  as  failure;
     that is, the number of confidence bounds violations in the preceding window
     of  observations met or exceeded a specified threshold. An example of using
     these RRRRAAss to graph confidence bounds and failures appears in rrdgraph.

     The SEASONAL and DEVSEASONAL RRRRAAss store the seasonal coefficients  for  the
     Holt-Winters  forecasting  algorithm  and  the seasonal deviations, respec‐
     tively.  There is one entry per observation time point in the seasonal  cy‐
     cle.  For  example, if primary data points are generated every five minutes
     and the seasonal cycle is 1 day, both SEASONAL and  DEVSEASONAL  will  have
     288 rows.

     In  order to simplify the creation for the novice user, in addition to sup‐
     porting explicit creation of the HWPREDICT, SEASONAL,  DEVPREDICT,  DEVSEA‐
     SONAL, and FAILURES RRRRAAss, the RRRRDDttooooll create command supports implicit cre‐
     ation of the other four when HWPREDICT is specified alone and the final ar‐
     gument _r_r_a_-_n_u_m is omitted.

     _r_o_w_s  specifies  the  length of the RRRRAA prior to wrap around. Remember that
     there is a one-to-one correspondence between primary data  points  and  en‐
     tries  in  these RRAs. For the HWPREDICT CF, _r_o_w_s should be larger than the
     _s_e_a_s_o_n_a_l _p_e_r_i_o_d. If the DEVPREDICT RRRRAA is implicitly created,  the  default
     number  of rows is the same as the HWPREDICT _r_o_w_s argument. If the FAILURES
     RRRRAA is implicitly created, _r_o_w_s will be set to the _s_e_a_s_o_n_a_l _p_e_r_i_o_d argument
     of the HWPREDICT RRRRAA. Of course, the RRRRDDttooooll _r_e_s_i_z_e command is available if
     these defaults are not sufficient and the creator wishes to avoid  explicit
     creations of the other specialized function RRRRAAss.

     _s_e_a_s_o_n_a_l  _p_e_r_i_o_d  specifies the number of primary data points in a seasonal
     cycle. If SEASONAL and DEVSEASONAL are implicitly  created,  this  argument
     for those RRRRAAss is set automatically to the value specified by HWPREDICT. If
     they  are explicitly created, the creator should verify that all three _s_e_a_‐
     _s_o_n_a_l _p_e_r_i_o_d arguments agree.

     _a_l_p_h_a is the adaption parameter of the intercept (or baseline)  coefficient
     in the Holt-Winters forecasting algorithm. See rrdtool for a description of
     this  algorithm.  _a_l_p_h_a must lie between 0 and 1. A value closer to 1 means
     that more recent observations carry greater weight in predicting the  base‐
     line component of the forecast. A value closer to 0 means that past history
     carries greater weight in predicting the baseline component.

     _b_e_t_a  is  the adaption parameter of the slope (or linear trend) coefficient
     in the Holt-Winters forecasting algorithm. _b_e_t_a must lie between  0  and  1
     and  plays  the  same  role  as  _a_l_p_h_a with respect to the predicted linear
     trend.

     _g_a_m_m_a is the  adaption  parameter  of  the  seasonal  coefficients  in  the
     Holt-Winters forecasting algorithm (HWPREDICT) or the adaption parameter in
     the  exponential  smoothing  update of the seasonal deviations. It must lie
     between 0 and 1. If the SEASONAL and DEVSEASONAL RRRRAAss are  created  implic‐
     itly, they will both have the same value for _g_a_m_m_a: the value specified for
     the HWPREDICT _a_l_p_h_a argument. Note that because there is one seasonal coef‐
     ficient  (or  deviation) for each time point during the seasonal cycle, the
     adaptation rate is much slower than the baseline. Each seasonal coefficient
     is only updated (or adapts) when the observed value occurs at the offset in
     the seasonal cycle corresponding to that coefficient.

     If SEASONAL and DEVSEASONAL RRRRAAss are created explicitly, _g_a_m_m_a need not  be
     the same for both. Note that _g_a_m_m_a can also be changed via the RRRRDDttooooll _t_u_n_e
     command.

     _s_m_o_o_t_h_i_n_g_-_w_i_n_d_o_w specifies the fraction of a season that should be averaged
     around each point. By default, the value of _s_m_o_o_t_h_i_n_g_-_w_i_n_d_o_w is 0.05, which
     means  each value in SEASONAL and DEVSEASONAL will be occasionally replaced
     by averaging it with its (_s_e_a_s_o_n_a_l _p_e_r_i_o_d*0.05) nearest neighbors.  Setting
     _s_m_o_o_t_h_i_n_g_-_w_i_n_d_o_w to zero will disable the  running-average  smoother  alto‐
     gether.

     _r_r_a_-_n_u_m  provides the links between related RRRRAAss. If HWPREDICT is specified
     alone and the other RRRRAAss are created implicitly, then there is no  need  to
     worry  about  this argument. If RRRRAAss are created explicitly, then carefully
     pay attention to this argument. For each RRRRAA which includes this  argument,
     there  is  a dependency between that RRRRAA and another RRRRAA. The _r_r_a_-_n_u_m argu‐
     ment is the 1-based index in the order of RRRRAA creation (that is, the  order
     they  appear in the _c_r_e_a_t_e command). The dependent RRRRAA for each RRRRAA requir‐
     ing the _r_r_a_-_n_u_m argument is listed here:

     •   HWPREDICT _r_r_a_-_n_u_m is the index of the SEASONAL RRRRAA.

     •   SEASONAL _r_r_a_-_n_u_m is the index of the HWPREDICT RRRRAA.

     •   DEVPREDICT _r_r_a_-_n_u_m is the index of the DEVSEASONAL RRRRAA.

     •   DEVSEASONAL _r_r_a_-_n_u_m is the index of the HWPREDICT RRRRAA.

     •   FAILURES _r_r_a_-_n_u_m is the index of the DEVSEASONAL RRRRAA.

     _t_h_r_e_s_h_o_l_d is the minimum number of violations (observed values outside  the
     confidence bounds) within a window that constitutes a failure. If the FAIL‐
     URES RRRRAA is implicitly created, the default value is 7.

     _w_i_n_d_o_w  _l_e_n_g_t_h is the number of time points in the window. Specify an inte‐
     ger greater than or equal to the threshold and less than or  equal  to  28.
     The  time  interval  this window represents depends on the interval between
     primary data points. If the FAILURES RRRRAA is implicitly created, the default
     value is 9.

SSTTEEPP,, HHEEAARRTTBBEEAATT,, aanndd RRoowwss AAss DDuurraattiioonnss
     Traditionally RRDtool specified PDP intervals in seconds,  and  most  other
     values  as  either  seconds or PDP counts.  This made the specification for
     databases rather opaque; for example

      rrdtool create power.rrd \
        --start now-2h --step 1 \
        DS:watts:GAUGE:300:0:24000 \
        RRA:AVERAGE:0.5:1:864000 \
        RRA:AVERAGE:0.5:60:129600 \
        RRA:AVERAGE:0.5:3600:13392 \
        RRA:AVERAGE:0.5:86400:3660

     creates a database of power values collected once per second, with  a  five
     minute  (300  second)  heartbeat, and four RRRRAAs: ten days of one second, 90
     days of one minute, 18 months of one hour, and ten years of one  day  aver‐
     ages.

     Step,  heartbeat,  and  PDP  counts and rows may also be specified as dura‐
     tions, which are positive integers  with  a  single-character  suffix  that
     specifies  a scaling factor.  See "rrd_scaled_duration" in librrd for scale
     factors of the  supported  suffixes:  "s"  (seconds),  "m"  (minutes),  "h"
     (hours), "d" (days), "w" (weeks), "M" (months), and "y" (years).

     Scaled  step and heartbeat values (which are natively durations in seconds)
     are used directly, while consolidation function row arguments  are  divided
     by their step to produce the number of rows.

     With this feature the same specification as above can be written as:

      rrdtool create power.rrd \
        --start now-2h --step 1s \
        DS:watts:GAUGE:5m:0:24000 \
        RRA:AVERAGE:0.5:1s:10d \
        RRA:AVERAGE:0.5:1m:90d \
        RRA:AVERAGE:0.5:1h:18M \
        RRA:AVERAGE:0.5:1d:10y

TThhee HHEEAARRTTBBEEAATT aanndd tthhee SSTTEEPP
     Here  is an explanation by Don Baarda on the inner workings of RRDtool.  It
     may help you to sort out why all this *UNKNOWN* data is popping up in  your
     databases:

     RRDtool  gets  fed samples/updates at arbitrary times. From these it builds
     Primary Data Points (PDPs) on every "step" interval. The PDPs are then  ac‐
     cumulated into the RRAs.

     The "heartbeat" defines the maximum acceptable interval between samples/up‐
     dates.  If  the  interval between samples is less than "heartbeat", then an
     average rate is calculated and applied for that interval. If  the  interval
     between  samples  is  longer than "heartbeat", then that entire interval is
     considered "unknown". Note that there are other things that can make a sam‐
     ple interval "unknown", such as the rate exceeding limits, or a sample that
     was explicitly marked as unknown.

     The known rates during a PDP's "step" interval are used to calculate an av‐
     erage rate for that PDP. If the total "unknown" time accounts for more than
     hhaallff the "step", the entire PDP is marked as "unknown". This means  that  a
     mixture  of  known and "unknown" sample times in a single PDP "step" may or
     may not add up to enough "known" time to warrant a known PDP.

     The "heartbeat" can be short (unusual) or long (typical)  relative  to  the
     "step"  interval between PDPs. A short "heartbeat" means you require multi‐
     ple samples per PDP, and if you don't get them mark the PDP unknown. A long
     heartbeat can span multiple "steps", which means it is acceptable  to  have
     multiple  PDPs  calculated from a single sample. An extreme example of this
     might be a "step" of 5 minutes and a "heartbeat" of one day, in which  case
     a  single  sample every day will result in all the PDPs for that entire day
     period  being   set   to   the   same   average   rate.   _-_-   _D_o_n   _B_a_a_r_d_a
     _<_d_o_n_._b_a_a_r_d_a_@_b_a_e_s_y_s_t_e_m_s_._c_o_m_>

            time|
            axis|
      begin__|00|
             |01|
            u|02|----* sample1, restart "hb"-timer
            u|03|   /
            u|04|  /
            u|05| /
            u|06|/     "hbt" expired
            u|07|
             |08|----* sample2, restart "hb"
             |09|   /
             |10|  /
            u|11|----* sample3, restart "hb"
            u|12|   /
            u|13|  /
      step1_u|14| /
            u|15|/     "swt" expired
            u|16|
             |17|----* sample4, restart "hb", create "pdp" for step1 =
             |18|   /  = unknown due to 10 "u" labeled secs > 0.5 * step
             |19|  /
             |20| /
             |21|----* sample5, restart "hb"
             |22|   /
             |23|  /
             |24|----* sample6, restart "hb"
             |25|   /
             |26|  /
             |27|----* sample7, restart "hb"
      step2__|28|   /
             |22|  /
             |23|----* sample8, restart "hb", create "pdp" for step1, create "cdp"
             |24|   /
             |25|  /

     graphics by _v_l_a_d_i_m_i_r_._l_a_v_r_o_v_@_d_e_s_y_._d_e.

HHOOWW TTOO MMEEAASSUURREE
     Here are a few hints on how to measure:

     Temperature
         Usually  you  have  some type of meter you can read to get the tempera‐
         ture.  The temperature is not really connected with a  time.  The  only
         connection  is that the temperature reading happened at a certain time.
         You can use the GGAAUUGGEE data source type  for  this.  RRDtool  will  then
         record your reading together with the time.

     Mail Messages
         Assume you have a method to count the number of messages transported by
         your  mail  server in a certain amount of time, giving you data like '5
         messages in the last 65 seconds'. If you look at the count of 5 like an
         AABBSSOOLLUUTTEE data type you can simply update the RRD with the number 5  and
         the  end  time  of your monitoring period. RRDtool will then record the
         number of messages per second. If at some later stage you want to  know
         the  number  of  messages transported in a day, you can get the average
         messages per second from RRDtool for the day in question  and  multiply
         this  number  with  the number of seconds in a day. Because all math is
         run with Doubles, the precision should be acceptable.

     It's always a Rate
         RRDtool stores rates in amount/second for  COUNTER,  DERIVE,  DCOUNTER,
         DDERIVE and ABSOLUTE data.  When you plot the data, you will get on the
         y  axis  amount/second  which you might be tempted to convert to an ab‐
         solute amount by multiplying by the delta-time between the points. RRD‐
         tool plots continuous data, and as such is not appropriate for plotting
         absolute amounts as for example "total bytes" sent and  received  in  a
         router.  What  you  probably  want  is plot rates that you can scale to
         bytes/hour, for example, or plot absolute  amounts  with  another  tool
         that  draws  bar-plots,  where  the delta-time is clear on the plot for
         each point (such that when you read the graph you see for example GB on
         the y axis, days on the x axis and one bar for each day).

EEXXAAMMPPLLEE
      rrdtool create temperature.rrd --step 300 \
       DS:temp:GAUGE:600:-273:5000 \
       RRA:AVERAGE:0.5:1:1200 \
       RRA:MIN:0.5:12:2400 \
       RRA:MAX:0.5:12:2400 \
       RRA:AVERAGE:0.5:12:2400

     This sets up an RRRRDD called _t_e_m_p_e_r_a_t_u_r_e_._r_r_d which  accepts  one  temperature
     value  every 300 seconds. If no new data is supplied for more than 600 sec‐
     onds, the temperature becomes _*_U_N_K_N_O_W_N_*.  The minimum acceptable  value  is
     -273 and the maximum is 5'000.

     A  few  archive  areas  are also defined. The first stores the temperatures
     supplied for 100 hours (1'200 * 300 seconds = 100 hours).  The  second  RRA
     stores the minimum temperature recorded over every hour (12 * 300 seconds =
     1  hour), for 100 days (2'400 hours). The third and the fourth RRA's do the
     same for the maximum and average temperature, respectively.

EEXXAAMMPPLLEE 22
      rrdtool create monitor.rrd --step 300        \
        DS:ifOutOctets:COUNTER:1800:0:125000000    \
        RRA:AVERAGE:0.5:1:2016                     \
        RRA:HWPREDICT:1440:0.1:0.0035:288

     This example is a monitor of a router interface. The first RRRRAA  tracks  the
     traffic  flow in octets; the second RRRRAA generates the specialized functions
     RRRRAAss for aberrant behavior detection. Note that the _r_r_a_-_n_u_m argument of HW‐
     PREDICT is missing, so the other RRRRAAss will implicitly be created  with  de‐
     fault parameter values. In this example, the forecasting algorithm baseline
     adapts quickly; in fact the most recent one hour of observations (each at 5
     minute  intervals)  accounts for 75% of the baseline prediction. The linear
     trend forecast adapts much more slowly. Observations made during  the  last
     day  (at  288  observations  per day) account for only 65% of the predicted
     linear trend. Note: these computations rely  on  an  exponential  smoothing
     formula  described in the LISA 2000 paper.  The maximum of 125000000 octets
     per second corresponds to a 1 Gbit/s link.

     The seasonal cycle is one day (288 data points at  300  second  intervals),
     and  the  seasonal adaption parameter will be set to 0.1. The RRD file will
     store 5 days (1'440 data points) of forecasts and deviation predictions be‐
     fore wrap around. The file will store 1 day (a seasonal cycle) of 0-1 indi‐
     cators in the FAILURES RRRRAA.

     The same RRD file and RRRRAAss are created with the  following  command,  which
     explicitly  creates  all  specialized function RRRRAAss using "STEP, HEARTBEAT,
     and Rows As Durations".

      rrdtool create monitor.rrd --step 5m \
        DS:ifOutOctets:COUNTER:30m:0:125000000 \
        RRA:AVERAGE:0.5:1:2016 \
        RRA:HWPREDICT:5d:0.1:0.0035:1d:3 \
        RRA:SEASONAL:1d:0.1:2 \
        RRA:DEVSEASONAL:1d:0.1:2 \
        RRA:DEVPREDICT:5d:5 \
        RRA:FAILURES:1d:7:9:5

     Of course, explicit creation need not replicate implicit create,  a  number
     of arguments could be changed.

EEXXAAMMPPLLEE 33
      rrdtool create proxy.rrd --step 300 \
        DS:Requests:DERIVE:1800:0:U  \
        DS:Duration:DERIVE:1800:0:U  \
        DS:AvgReqDur:COMPUTE:Duration,Requests,0,EQ,1,Requests,IF,/ \
        RRA:AVERAGE:0.5:1:2016

     This example is monitoring the average request duration during each 300 sec
     interval  for  requests  processed  by a web proxy during the interval.  In
     this case, the proxy exposes two counters, the number of requests processed
     since boot and the total cumulative duration  of  all  processed  requests.
     Clearly  these counters both have some rollover point, but using the DERIVE
     data source also handles the reset  that  occurs  when  the  web  proxy  is
     stopped and restarted.

     In  the RRRRDD, the first data source stores the requests per second rate dur‐
     ing the interval. The second data source stores the total duration  of  all
     requests  processed  during  the  interval divided by 300. The COMPUTE data
     source divides each PDP of the AccumDuration by the  corresponding  PDP  of
     TotalRequests and stores the average request duration. The remainder of the
     RPN expression handles the divide by zero case.

SSEECCUURRIITTYY
     Note  that  new  rrd files will have the permission 0644 regardless of your
     umask setting. If a file with the same name previously exists, its  permis‐
     sion settings will be copied to the new file.

AAUUTTHHOORRSS
     Tobias Oetiker <tobi@oetiker.ch>, Peter Stamfest <peter@stamfest.at>

1.10.0                             2026-05-23                       _R_R_D_C_R_E_A_T_E(1)
